DG4000 - a firmware investigation

[fact]

It's been quiet the last couple of weeks. I'm hoping someone is looking into this.
I tried to obtain a copy of previous work of cyber to investigate myself. So far I did not get anything.

I'll just wait until someone comes up with a solution or stuff I can work on myself. In the mean time I'm maxing out at 60MHz.

[Pasky]

Just got my function generator today.  It had 01.07 on it and just modified it to a DG4202.  Thanks for all your work!

[Pasky]

Quick question, I re-calibrated the generator using the measure value/input value and saving on the calibration points.  I noticed the Square wave at 50Mhz looks almost like a Sine wave and hardly resembles a square wave.  Is this normal or is my calibration off?

EDIT:

Nevermind, didn't adjust the period 

[TooOldForThis]

Quick question, I re-calibrated the generator using the measure value/input value and saving on the calibration points.  I noticed the Square wave at 50Mhz looks almost like a Sine wave and hardly resembles a square wave.  Is this normal or is my calibration off?

Check this service note from Rigol entitled "Why does my scope show a Sine Wave when I expect a Square?" http://beyondmeasure.rigoltech.com/acton/attachment/1579/f-00ca/0/-/-/-/-/file.pdf

[Pasky]

Thanks for that!

[Sparky]

New firmware: DG4000 Series v00.01.10.zip (uses bootloader v00.00.06)

Edit: The bootloader (00.06) is the same as was released with the 01.09 firmware update.

[TooOldForThis]

I just loaded 1.10.   Lord knows what this version fixed, but it wasn't the counter function.  At low frequencies, a period measurement read from the SCPI interface is still 13nS longer than the same measurement displayed on the LCD.  The SCPI and LCD frequency values don't match either.  Neither the SCPI or LCD frequency values are the reciprocals of either the SCPI period or LCD period values.  They've done a lovely job of maximizing the number of conflicting values reported for a single measurement.

[Rory]

I just loaded 1.10.

Extended bandwidth mod still working?

[plasijo]

Dear all,
may I ask you, where is the generator available for downloading? Unfortunately bayfiles.net is not working.
Thanks.

[Chainsaw_]

I just loaded 1.10.

Extended bandwidth mod still working?

Retained its new-found DG4162 identity fine, yes. Chassis label is DG4102 and mod done on 1.07; do you want me to sweep it to be sure?

[Chainsaw_]

where is the generator available for downloading?

Page seven, bottom post, from bandgap.

[Teneyes]

Here is an interesting comparison with $1200 unit testing

[Rory]

I just loaded 1.10.

Extended bandwidth mod still working?

Retained its new-found DG4162 identity fine, yes. Chassis label is DG4102 and mod done on 1.07; do you want me to sweep it to be sure?

Mine reverted back to DG4062 from DG4202 when 1.10 installed.

[DrMag]

Retained its new-found DG4162 identity fine, yes. Chassis label is DG4102 and mod done on 1.07; do you want me to sweep it to be sure?

Mine reverted back to DG4062 from DG4202 when 1.10 installed.

Rory, was it possible to reenable the added bandwidth after the update?

Can anyone else verify/confirm one of the two conflicting results reported here?

[Rory]

It looks like Chainsaw changed from 1.07 to 1.10  and Rory changed from 1.09 to 1.10, that could be the difference
Not to worry , the Man is working on it , see sniffing

It was at 1.08 before the update. Dang, I wish I'd changed it to 4162 before updating the firmware.

[Sparky]

DG4000 Calibration Restoration:

I just finished calibrating my DG4000 for Channel 1 and 2, and the frequency response is flat up to 200 MHz and well within better than +/- 0.8 dB.

So what is calibration? I have concluded that it is removing the un-calibration that you end up with after you install the firmware patch to extend the frequency!

So to calibrate (or rather restore the previous calibration of) your DG4000:  Press - Utility, Test/Cal, Secure Code, (put in PW) 2010, Enter, Secure OFF, (the result should be) Secure ON, Access gained.
Then for Cal. select HFLAT, Call Point (start at HFLAT 1-1), Measure Value, Input Value, press Save and wait for it to complete (about 3 seconds), select Cal Point for the next step.  Repeat for all 60 steps.  Then your done and you didn't have to connect anything up to your DG4000, except for AC power.

I was using my DG4062 (now DG4162) and decide to check for amplitude flatness up to 150 MHz.  I found it to be quite bad above 60MHz; large +ve and -ve variation in amplitude with frequency.  It is much worse than Teneye's shows here or here.  So I thought I hadn't done the "calibration restoration" correct as per ted572's instructions.

I repeated the "Calibration Restoration" for Amplitude, Low Freq Flat and High Freq Flat and Saved the results after completing the Measure Value/Input Value step for each set of points.  When I repeat the frequency sweep there doesn't seem to be any noticeable improvement in amplitude, compared to before restoring the calibration.  If I go back into calibration to inspect values I find the column of measured values ("MeasValue") is empty for all Amplitude, Low Freq. Flat, etc.  It doesn't show the values that were previously restored by the "Measure Value", "Input Value" button presses.  Are other people's experiences the same?  I'm thinking the calibration values are not being saved, and that's why I don't see any effect of the calibration.

[Sparky]

We shouldn't have to do anything with the Calibration for a DG4162.  This is only required for  the DG4202 (200MHz hack). There are also a couple of other small glitches we can run into with the DG4202 modification.  Which is why I now suggest staying away from the DG4202 hack.  And you can also loose your hack and end up back with a DG4062 by installing a recent Firmware release update.  And then end up SOL for going back to even a DG4162 (the preferred model).

How I would measure the DG4162 output level frequency response:  I would set the DG4162 for 50 Ohm output impedance (preferred for these higher frequencies for a flat output into your 50 Ohm Input device).  If I used an O'Scope then I would Terminate its input with 50 Ohms (either internally, or with a 50 Ohm Feed Through Termination).  The measurement should be into a 50 Ohm measurement device using 50 Ohm Coax.  A RF Power Meter, Spectrum Analyzer, etc. is preferred over a O'Scope for better output level measurement accuracy.  Although many hobbyist may not have this type of equipment, and therefore an O'Scope may be the their only choice, and should be fine if it has at least a 200MHz BW.

Edit:  If a 50 Ohm Feed Through Termination for the O'Scope isn't available, a BNC Tee Adapter (one Male port and two Female ports) with a BNC 50 Ohm Termination on one of the BNC Tee's Female ports can be used.  If a 50 Ohm Termination isn't available, a 50 Ohm resistor can be soldered to one of its Female ports (with very short leads).  If a BNC Tee Adapter isn't available, I believe that they can generally be found at a Radio Shack store.   

Thanks ted572 for your comments!  My DG4062 was modified to DG4202 for a while, but the recent DSP v01.10 firmware wouldn't install on my unit (possibly I was still using the bootloader before v01.06 which was released with the previous DSP v01.09 firmware), and so I ended up downgrading to DSP v01.04, then upgrading to DSP v01.06, applied the DG4162 model, and then upgrading again to DSP v01.08.

Anyways, I appreciate your comments and from your notes my oversight was obvious --- I forgot to set the scope input to 50 Ohm impedance.  I repeat the measurements (600mVp-p sine wave, 1MHz to 159Mhz sweep over 10sec) now with 50 Ohm output on the DG4162, and the o'scope (MSO2072A modded to MSO2302A) input set to 50 Ohm internal impedance.  I used RG-58 A/U coax.

The results I obtained this time were much better, although the first was still much different than those published earlier so I repeated the frequency sweep with 3 different coax cables I have (different manufacturers).  I'm surprised to find significant differences with simple RG-58 A/U!  I have attached the screenshots from the o'scope --- you can easily see the 10sec over which the frequency sweep runs.

The last result seems very good.  I have to conclude no problem with the DG4162 calibration, and that lack of 50 ohm termination and coax quality was responsible for erroneous earlier results.  It is good to see things working correctly this time. 

Anyone feel free to leave some comments or post comparisons of frequency sweep from your DG4000 units.

Edit: Some extra info.  My AWG is DG4162 model (updated from DG4062), with following hardware/firmware revisions:
Software Version: 00.01.08.00.02
FPGA Version:     00.01.09.00
Hard Version:     01.01
Keyboard Version: 04.01

[Teneyes]

Anyone feel free to leave some comments or post comparisons of frequency sweep from your DG4000 units.

Here are sweeps on 3 different coax cables
The sweeps are linear from 10-160Mhz over 60sec.
I have to note which cable is best

Note: DSO was in single sweep with DSO trigger out connected to DG sweep trigger input

[Sparky]

Anyone feel free to leave some comments or post comparisons of frequency sweep from your DG4000 units.

Here are sweeps on 3 different coax cables
The sweeps are linear from 10-160Mhz over 60sec.
I have to note which cable is best

Note: DSO was in single sweep with DSO trigger out connect to DG sweep trigger input

Nice, Teneyes!  The last result looks especially good.  That was a good idea to use trigger out of the DSO to start the sweep on the DG.  I will put a note on my cables, too.

Out of curiosity: Does anyone know which part(s) of the cable (coax itself, BNC connectors, or termination quality) is responsible for non-flatness and roll-off at high frequency.  If its not the coax itself, does that mean someone could "repair" a poor cable by replacing the BNC connectors?

[Sparky]

Sparky:   I generally avoid using RG-58.  I use RG-223 in its place for a flexible coax cable for my test cables.  Other cables that make good test cable, although aren't as flexible, are RG-141, 400, etc. with PTFE (teflon).
I'm kind of surprised that your RG-58 is as poor as it is, but that could be due to it having been flexed a lot over the years, having been kinked, pinched, BNC terminations being flaky, etc.  To inspect the BNC (etc) connectors you would have to disassemble them from the cable ends are redress the cable by cutting an inch or so and reterminating  the connections.  Do NOT just stick them back together without having fresh and clean connections.  An important thing to keep in mind with coax is not to over flex it too much, and to never bend a kink in it, or pinch/crush it.  If you kink it and then straighten out 'even a flexible coax cable', the site of the kink will show up with a Time Domain Reflectometer test as a purtibation/discontinuity in the coax.

I suggest getting some new RG-223 and BNC connectors to use with it if you have any doubt about your coax test cables.  We should be cautious buying coax from a Hamfest flee market in the parking lot.  Does it look fresh, clean, and undamaged?  Then hey, it may be Ok (?).  And if people have any doubt on how to properly dress the cable for a proper connector termination, they should look it up and follow the recommended process.  Some hobbyist have a hard time assembling coax cables/connectors simply because they have never been taught or studied the proper methods.       Regards, Ted

Thanks for the info, Ted!  Very helpful!  I wouldn't mind buying the cable and connectors and assembling myself so I can make custom lengths as needed, and from a quick look on Pasternack it is cheaper to buy the parts.  I note there are clamp/solder BNC connectors, so no costly crimp tool needed.  Would you happen to have a recommendation of a tool for cutting the insulation and braid to the specific lengths as needed for the connectors?  Thank you!

[MiataMuc]

wild idea (from someone who has never used a jtag-device other than for reading his DS2000 firmware out.. ): would it be possible to put an older (=hack-compatible) Firmware-Version on the DG4000 by jtag?

Flo

[rf-loop]

Anyone feel free to leave some comments or post comparisons of frequency sweep from your DG4000 units.

Here are sweeps on 3 different coax cables
The sweeps are linear from 10-160Mhz over 60sec.
I have to note which cable is best

Note: DSO was in single sweep with DSO trigger out connected to DG sweep trigger input

Cables are different, of course. It can read from cables datasheets.

One question about this interesting thing.

Did you use exactly same cable lenght (and not only physical lenght but with around same travel time lenght because it matter. And images looks like there is signal travel time differencies between cables - and impedance mismatches)?

(this question is because there is not 50ohm source impedance and not 50ohm load impedance
Oscilloscope input is mostly NOT 50 ohm impedance at all. It is more like 50ohm resistance for DC.  There is around 50ohm resistance and then inductive and capasitive reactances series and parralle and because this scope input reflects back something more or less and depending cable lenght it affect more or less with different frequencies..)

[Teneyes]

Did you use exactly same cable lenght (and not only physical lenght but with around same travel time lenght because it matter. And images looks like there is signal travel time differencies between cables - and impedance mismatches)?

@rf-loop, Yes the DG4000 generator was set to 50ohm output  and there was a 50 ohm feed-thru termination on the DS2000. 
Yes 2 of the cables were 3 feet long and 1 was 6 feet long.
After I did the Cable sweep  test,  I did do a timing test  to see how much longer the 6' cable was  in time delay, so I connected the shorter cable on Chan.1 and the longer cable on Chan.2  and feed the Sync output pulse from each of the Gen 2 outputs (50 ohm) . One cable is about 6.4 nS longer. the Gen outputs were 'Frequency Locked'.
 
But as I show in the 2nd Pix the timing  drifts over time ( note clock 3hours, and my stored Ref traces, pink and white)  ,
When I reset the 'Phase Align'  the 2 pulses are aligned at the Gen again. see Pix #3 (5 hours)

That is something to know when locking 2 Channels on the DG4000. Frequency Lock does not Lock the phases also.
 

[dpenev]

Hi Gents,

The slope of my DG4062 square wave is about 10ns, and this is what i see spicified in the documentataion 
On your pictures I see quicker slopes (about 2ns).
Have you done some calibration to get this?

Regards
Dimitar

   

[Teneyes]

On your pictures I see quicker slopes (about 2ns).
Have you done some calibration to get this?

If you are referring to the step changes I just posted, please NOTE that I stated the traces are of the Sync outputs, fast switching pulse, but Fixed at 1.6 V .
Good for you to notice the slope, but did anyone notice the trigger point