Rigol DG4162 Amplitude Inaccuracies

[StubbornGreek]

I recently purchased a DG4162 and overall I do like the instrument but here's some FYI:

Here's what a 2Vpp sine wave looks like starting at 50MHz (where the signal displays just proud of the 2Vpp). At 65MHz, you'll notice the wave registers just under the 2Vpp then a bit lower at 66MHz (20mV lower). By 70MHz the error has become a bit gross, at 100MHz its displaying almost a 25% error and at 160MHz...

I'm not finding much else to criticize but jeez.

[Rerouter]

are you sure your scope is not contributing some of that loss of amplitude? you failed to mention the model of scope

[EV]

Are you using a 1:10 probe? Try direct BNC cable with 50 ohm feed thru terminator.

[StubbornGreek]

are you sure your scope is not contributing some of that loss of amplitude? you failed to mention the model of scope

Scope is a Rigol DS2202 but I get the same results on my Owon - using the second channel on either scope doesn't change anything either.

Are you using a 1:10 probe? Try direct BNC cable with 50 ohm feed thru terminator.

No probe, just a direct BNC cable (I tried more than one). The results are the same with the cable terminated at 50 Ohms and with a direct cable (no termination) and the FG set to High-Z.

[EV]

Here results with my DG4162 generator and DS2202 scope:

In the first picture is sweep from 1 MHz to 160 MHz using Rigol 50 ohm feed tru terminator. DG4162 output is set to 50 ohm. The change between max and min level is under 1 dB ( 12 % ) and compared to 1 MHz level the change is under 0.5 dB ( 6 % ).

In the second picture is the same sweep with T-Adapter and 50 ohm terminator. There is no change in the curve, if feed thru terminator is used.

[EV]

From Rigol DC4162 specs:

Amplitude Flatness (relative to 100 kHz, 1.25Vpp Sine wave, 50 ohm)
Typical
<10MHz: ±0.1dB
<60MHz: ±0.2dB
<100MHz: ±0.4dB
<160MHz: ±0.8dB

For comparison in the picture is sweep with HP8642A generator from 5 MHz to 160 MHz.

[kg4arn]

I just measured my DG4162 with  an Agilent DSOX3024A, 50 ohm input.
I have previously measured my scope's frequency response as -1.2 dB at 160 MHz and -3dB at 220MHz (that's probably with an error of +/- 0.5 dB)

The DG4162 sine wave output measures -0.9dB at 160MHz, which is within the error limits.
I get the same results with my vintage HP 8654A signal generator.

Not precision measurements, I admit, but they correlate and suggest that at least my DG4162 is performing well.

[BravoV]

Assuming the Rigol is not defective, crappy coax cable or bad terminator ? Try another set maybe ?

[David_AVD]

No probe, just a direct BNC cable (I tried more than one).

Is the cable rated for the frequencies in question?

[alm]

Don't expect accurate amplitude measurements from a scope. A scope is usually rated something like 2% at DC, and -3 dB (-30%) at its rated bandwidth. The accuracy between DC and its rated bandwidth can be anywhere in this range, it's only guaranteed to be less than -3 dB. I haven't seen flatness specs for the pass band, at least for low-end scopes.

[Rufus]

The DG4162 user guide specifies a maximum output of 1v pk-pk into 50 ohm at 160MHz. What you will get with a bit of coax feeding a 1M/15pf (or whatever) scope input is anyone's guess.

[jpb]

I am curious about this as I've been looking at AWG and AFGs and trying to compare specs which is difficult because they all state them at different levels.

One thing that I've noticed about the Rigol is that it avoids speccing its amplifier.

I think what Rigol does is rely on the DAC directly for signals up to 1Vpp (into 50ohms) all its specs are at 0dBm or about 0.3V.

It then has an amplifier with a fairly low frequency roll off giving x10 (or perhaps x8) up to 20MHz.

The gain is set to x4 below 4V (if you look at the manual where it talks about limits on the output it is 11.25V above 4Vpp which is 1.125V at the input of the amp if it is x10,
below 4V it is a limit of 4.5V which also equates to 1.125V at the input if the gain is x4).

The x4 is able to go to 5Vpp if the DAC is slightly overdriven to 25mA (20mA is the typical from the spec sheet) and this is specced up to 60MHz and then it is set at x2 up
to 120MHz and above this the amplifier is probably switched out or set to x1.

In the tear down what were described as switched attenuators I think are just switched feedback resistors to set the gain on the output amp.

It is interesting to note that gain flatness is specced for a 1.25V sine which may be the limit of the DAC so again the amplifier is not being specced at all.

What all this means is that if you keep to low signal levels below 1Vpp you will get much better results than going to 2Vpp or higher where the low frequency amplifier is needed
and there are no specs given.

The original poster posted results for 2Vpp where the amplifier is needed. Most of the responses above have been for signal levels at 1V or below where the amplifier is not switched in (if I'm correct).

It will be interesting to see peoples results for the higher amplitude settings. My suspicion is that Rigol is able to come in at a low cost on the 4000 series (even compared to its own 2000 series) because it is using a relatively cheap op-amp and setting different gain levels for the output. I would guess that much of the cost of more expensive units goes in the design of the wide bandwidth linear amplifier needed to change the 1Vpp DAC output up to 10Vpp that is specced for almost all such instruments.

[EV]

It will be interesting to see peoples results for the higher amplitude settings.

1 Vpp is max amplitude between 120 - 160 MHz for DG4162.

[jpb]

It will be interesting to see peoples results for the higher amplitude settings.

1 Vpp is max amplitude between 120 - 160 MHz for DG4162.

Yes, but a comparison could be made between 1Vpp up to 120MHz and 2Vpp up to 120MHz.

Amplitude makes quite a difference I think. I've also been looking at the TTi TG5011 which has separate specs for <1V and > 1V. Also at the Tabor WW5061 which in early specs used the level of 5V (it has been around a while and older spec sheets are available on the web) and then reduced it to 3Vpp and improved the high frequency distortion figures as a result.

For example, the original Tabor WW5061 spec for Harmonic Distortion up to 25MHz was given at 5Vpp and the figure was -28dBc. The latest figure, for the same hardware, is -40dBc but this is given for 3Vpp. Then you are supposed to compare this with the Rigol which is -50dBc all the way up to 100MHz but it is given at 0dBm which is only 0.31Vpp or one tenth of the amplitude.

I've found myself going round in circles trying to do a fair comparison between the Rigol DG4062, the TTi TG5011 and the Tabor WW5061 all of which I can get for about the same price (the Rigol new and the other two as ex-demo or old stock as their new list prices are much higher).

[EV]

Yes, but a comparison could be made between 1Vpp up to 120MHz and 2Vpp up to 120MHz.

Here are 1...120 MHz sweeps for 1 and 2 Vpp with DG4162 as you wanted.

[StubbornGreek]

At 160MHz and 1Vpp, things don't look much better:

[EV]

At 160MHz and 1Vpp, things don't look much better:

You have connected generator to scope with BNC cable and generator output is at HiRes mode? Change generator output to 50 ohm output mode and use 50 ohm terminator (with T-adapter) at scope input. You will get more reasonable results!

[EV]

Generators HiRes mode is for 1:10 probe input!

[kg4arn]

Yes, but a comparison could be made between 1Vpp up to 120MHz and 2Vpp up to 120MHz.

....

Sweeping from 1 MHz to 120 MHz at both 1Vpp and 2Vpp I can see no significant change in the output amplitude into a 50 ohm load.

[jpb]

Yes, but a comparison could be made between 1Vpp up to 120MHz and 2Vpp up to 120MHz.

Here are 1...120 MHz sweeps for 1 and 2 Vpp with DG4162 as you wanted.

Thanks for that. I notice though that the scope is set for 500k samples per second, at least that is what it says at the top of the screen shot, this means it will be undersampling on frequencies above 250kHz so this may do odd things to the magnitude. I presume the sample rate is set by the time the sweep takes and the amount of memory available so it may be necessary to set different frequencies rather than doing a sweep.

Anyway I don't want to waste your time with multiple experiments.

[AndyC_772]

Set the scope to peak detect acq mode and it'll run the ADC flat out all the time.

[Rufus]

I think what Rigol does is rely on the DAC directly for signals up to 1Vpp (into 50ohms) all its specs are at 0dBm or about 0.3V.

It then has an amplifier with a fairly low frequency roll off giving x10 (or perhaps x8) up to 20MHz.

I haven't attempted to look at or understand the DG4162 circuitry but my initial assumption was it has DAC good for 160MHz sinewaves and an output stage that is only good for around 50MHz. It only does sinewave above 50MHz and they just let the output stage loose 10db at 160MHz. It will all be calibrated in software so it should be accurate, just the higher frequency you go the less amplitude you can have.

I would also be surprised if the hardware for the 3 models isn't identical.

[jpb]

Yes, but a comparison could be made between 1Vpp up to 120MHz and 2Vpp up to 120MHz.

....

Sweeping from 1 MHz to 120 MHz at both 1Vpp and 2Vpp I can see no significant change in the output amplitude into a 50 ohm load.

That is a good result. But it may be that some examples are better than others given that others seem to have worse results. If Rigol can get consistently good results at higher amplitudes they should publicize them on the specifications. By only giving results at 0dBm (or 1.25V for amplitude flatness) it leaves the potential buyer with no idea as to performance at typical amplitudes (given that most of these instruments from Agilent, TTi, Tektronix, Tabor and so on have 10Vpp as a maximum I assume that typical waveforms would be in the range 3 to 5Vpp rather than 0.3Vpp).

I don't want to knock Rigol who have given the user the option to trade off magnitude against frequency, and many users are probably happy with small signal levels if it allows them to go to much higher frequencies. It just seems to be a general problem with specifications is that each manufacturer chooses their own amplitudes, frequency ranges and so on to give figures and then it is very hard to compare. If you're a company you can get demo machines and do your own tests. If you're buying as an individual (like me) you have to rely on spec sheets and forums like this one.

[Rufus]

Generators HiRes mode is for 1:10 probe input!

At 160MHz a typical x10 scope probe has an impedance of around 70 ohms - hardly 'high' so it will be about 3db down fed from a 50 ohm source.

I think the OP needs to get a bit of an rf clue.

[jpb]

I haven't attempted to look at or understand the DG4162 circuitry but my initial assumption was it has DAC good for 160MHz sinewaves and an output stage that is only good for around 50MHz. It only does sinewave above 50MHz and they just let the output stage loose 10db at 160MHz. It will all be calibrated in software so it should be accurate, just the higher frequency you go the less amplitude you can have.

I would also be surprised if the hardware for the 3 models isn't identical.

Looking at the photos on this forum (in the review and also the video tear-down) there seems to be three relay switchers by the output amplifiers which I think actually switch gains (2x 4x and 10x) - this seems to fit in with the manual stating that there are different back voltages allowed on the ports depending on whether or not amplitude is above or below 4Vpp and frequency below 10kHz. The 2x gain is only needed well above 10kHz so is not mentioned in the manual.

There is a fourth relay which I think switches in or out a final filter stage and I guess this is what determines the rise time of the square waves for instance.

I think the hardware of the models is identical except perhaps for this last filter stage. The fact that they all have identical rise times for arbitrary waveforms indicates that this is the case - the arbitrary wave forms do not pass through this final filter (there is an earlier image suppression 160MHz low pass filter that they all have).

The software could compensate for the gain dropping with frequency but could only be for typical values so some devices may be incorrectly compensated (I'm just surmising).