Rigol DS1000Z-S / DS2000A-S internal AWG Calibration?

[TurboTom]

Can anyone provide some information on how the internal arbitrary waveform generator of the more mature Rigol Oscilloscopes is supposed to be calibrated? Several years ago, I acquired a DS2072A-S from Rigol Europe's Clearance Center. Since for what ever reason it contained the MSO hardware, I turned it into an MSO2302A-S. Only later I found out that the AWG section apparently has got a problem with a DC offset of about -150mV if the output level is increased above 799mVpp 50Ohms / 1.59Vpp / hi-Z on both channels (ch1 is a little worse than ch2).

Recently, I had a closer look at the AWG board to analyze if the offset may be a hardware problem, but while I found some peculiarities, it apparently isn't. FYI, here are photos of the really tiny AWG module and some information on its technical details:






Main semiconductors:
FPGA Altera Cyclone IV EP4CE22F17C8N
D/A Converters AD9744
Aux D/A Converter DAC8560
Transimpedance amps / pre-drivers: AD8009
Output amps: LMH6702
Offset voltage amp: AD8512

First thing that I noticed was the somewhat "roasted" appearance of ch2's 50R output resistor. Even though it must have been overloaded at some time, it's resistance is still spot-on and the output amp is also performing well. The output overvoltage protection (LM339) should have taken care of that but maybe some too low voltage to trigger the protection had been applied to the output for considerable time so the 1210 resistor just overheated somewhat. This must have happened before I received the instruement, probably while it spent its "first life" as a demo unit...

Moreover, I found an edge of the LM339 chipped, which happened probably during assembly since there's a threaded stud located next to it. This shouldn't really matter, though...

The final output amps are only utilized at output voltages above ~1.6Vpp (high-Z) or ~800mVpp (50Ohm Load), otherwise they are bypassed by the second relays in the rows. The rightmost relays enable the generator outputs while the leftmost ones select between a 20dB / 63.7 Ohm attenuator or a straight-through. The pre-drivers's output series resistance is 63.2 ohms and thus matches the attenuator. And this is the interesting part: The offset is generated by a separate 16 bit D/A converter, amplified by an opamp to quite a high level and then resistively added to the main output signal via the 249 Ohms resistors located on the bottom of the PCB. Since the load "sees" the 63 Ohms and the 249 ohms in parallel, the effective output resistance is ~50 Ohms.

This way, it's possible to add a decent range of offset without having to sacrifice resolution of the AC output signal. On the down side, the offset output amp has to be run at quite a considerable operating voltage, in this case +-15V. The offset voltages and two more internal DC control voltages are generated via a single-channel DAC8560 D/A converter, and sampled via HC4051 analog multiplexer to four channels to chare small storage caps, connected to a TLC274 CMOS amplifier that acts as a buffer.

Altogether, the design is quite straight-forward and the compopnent choice is decent. Since there's no non-volatile memory present on the AWG module, all the calibration has to be done via and stored in the host oscilloscope. If I only knew how that's supposed to be accessed... . Any help would be greatly appreciated .

Cheers,
Thomas

[mawyatt]

Interesting, and nice discussion on the design!!

-150mv is quite large an offset to deal with, hopefully you get at this with the cal routine. If not, maybe a hardware "tweak" with a added resistor

Best, 

[tv84]

Don't know if these help in any way...

[TurboTom]

Thanks to all the contributors -- unfortunately, I already checked the available documentation without any success. I guess the "hidden" SCPI commands will somehow permit to calibrate the unit, but unless the syntax and parameters are known, I will have little success. I thought there may actually be something like an "extended self-cal" that could be run after the normal self-cal that would use the scope itself to calibrate the AWG by connecting its outputs directly to the scope's input channels. But it appears that that's not the case.

To reason from the strange configuration that I found the scope to be in when I received it (MSO main PCB with AWG board in a DS2072A-S casing), and the fact that I bought it as a clearance item, it's well possible that some technician at Rigol Europe just assembled a unit from components of several defective units without accomplishing a full calibration afterwards.

Whatsoever, it's performing well otherwise and I've got more than enough (stand-alone) AWGs that the offset is more a "luxury problem" than any real issue. It's just that I like to have my equipment in a properly working shape... And if the cal scheme of this unit was known, who knows, maybe this information may also help to calibrate the DG800/900/2000 series... 

[hexreader]

I wonder if the 16-pin chip next to the FPGA is a serial configuration memory for the FPGA?

Possibly (possibly not) calibration info is held there?  Must admit, I struggle to imagine that it would be.

Cannot find any info on the chip, but I am curious about it.

[TurboTom]

No, that's just a 74AHC595 shift register...

[mawyatt]

If both channel outputs show the offset that kinda rules out the separate in-line amplifiers and AD9744 and points to something common to both channels which apparently is the AD8512, TLC274, HC4051 and DAC8560. If the offset is stored in caps from the HC4051 leakage could be the culprit?

Best,

[TurboTom]

I don't think this is actually the problem. At those output levels where the output drivers (LMH6702) are bypassed, there isn't any noticable offset, and moreover, the two offset ADC channels stay at zero. Only if the preset output amplitude invokes enabling the drivers, there is some small negative voltage supplied by the offset ADC channels. As per the datasheet for the LMH6702, the non-inverting input supplies some microamps of current which I measured on both channels and which is within spec. I assume that the negative offset ADC output is meant to compensate this current to keep the output of the drivers at zero volts. It's well possible that the scope had been calibrated for a different AWG assembly and this has not been touched after the assemblies had been swapped. I'm almost 100% certain that this is a calibration issue.

It's so sad that today's ascending companies don't publish the amount of service information that the "T&M Dinosaurs" did some three decades ago... Well, nowadays, they don't do this anymore as well. 

[mawyatt]

If I understand, the offset is likely to compensate for the bias input currents of the current mode feedback output driver amps and thus a calibration error.

Totally agree, the old Tektronix and HP manuals are priceless. Without them I wouldn't have been able to rebuild and repair a pair of Tek 2465 analog scopes and a pair of HP/Agilent 34401A DMMs.

Best,