The Siglent SDG2042X Thread

[Messtechniker]

I don't think its a ground loop issue. Why should the 50 Hz noise
disappear when switching the generator output signal off via its output button?
Disconnecting the BNC cable from the generator output also makes the 50 Hz noise disappear.
Same result as depicted for the SDG 2042X in the OFF state.
This excludes the possibility of a ground loop on the measurement side.
Maybe someone would like to do the same measurements on his SDG 2042X unit.

[blurpy]

Maybe someone would like to do the same measurements on his SDG 2042X unit.

Did you see my tests just a few posts back?

[TurboTom]

@Messtechniker:

You may want to perform the same experiment with a 50 Ohms through-terminator connected directly to the AWG's output BNC. This way, even if powered off, there's always a (fairly) low impedance link of the signal conductor to ground close to the source. If you now notice the 50Hz hum all the time the AWG is connected to the power line (regardless wether turned on or off), it's clearly a ground loop. If it's only there while the output is active, it's the AWG's fault.

[Messtechniker]

Solved: Its a ground loop because the 0V
reference potential at the SDG output is internally connected to
the protective ground conductor. As the PC 0V is also 
connected to protective ground --->perfect ground loop
Only (safe) way out. Use a balanced input.

Thanks TurboTom for putting me on the right track

[Johnny B Good]

Solved: Its a ground loop because the 0V
reference potential at the SDG output is internally connected to
the protective ground conductor. As the PC 0V is also 
connected to protective ground --->perfect ground loop
Only (safe) way out. Use a balanced input.

Thanks TurboTom for putting me on the right track

 Apologies for the seemingly off-topic contribution that follows but in view of the fact that many of the members following this thread will have owned or still own one of these Feeltech AWGs or its many clones and copycats, I think it will still provide an interesting insight into this issue of unwanted mains earth ground loops arising out of misguided attempts by both users and manufacturer alike to eliminate the half live mains ESD issue in what is essentially a class II electrical appliance. Yes! It is a bit of a rant against Feeltech in particular but considering the circumstances, this can't really be avoided.

 This discussion over the effect of mains earth ground loops in such test configurations reminds me of the singular charm of FeelTech's FY6600 "Toy AWG" lack of a safety earth connection which neatly disposed of this particular pitfall. Admittedly, this did present a half live mains ESD issue if you forgot the mantra "Connect ground return first and disconnect it last" when connecting up to an ESD vulnerable part of an otherwise normally ESD proof device.

 Unfortunately, Feeltech had taken note of the many complaints about this half mains touch voltage issue and the resulting ESD hazard made in the long running EEVBlog FY6600 topic thread so when they decided to release a revamped FY6600 in the form of the FY6800, the chief bean counter bastard in charge of the whole Feeltech operation must have insisted that the C13 socket upgrade be wired up at the lowest possible cost regardless of the consequences.

 In this case, the consequences were the creation of an earth loop issue where none had previously existed, an increased noise ripple level in the psu's zero volt return rail and a reduction in electrical safety since they'd simply stolen one the two zero volt return wires in the 6 way ribbon cable connection between the psu and main board so as to divert it directly to the PE tag on the C13 socket as cheaply as possible.

 Not only did this wilful act of vandalism introduce an earth loop issue that had never previously existed in the FY6600 and its predecessors, it had actually made it less safe than before since the stolen wire lacked sufficient surge current rating to handle the hefty currents that could arise from a live to ground contact fault!

 Incredibly, the FY6900 has also had this earthing bodge inflicted upon it. You'd have thought by then, that they would have at least come up with a safer way to connect the BNC ground rail to the C14 socket's PE tag seeing as how the component parts had all been reassembled into a completely rejigged plastic enclosure. What made this bodge even worse than the one used in the FY6800 was the routing of this wire against components on the smpsu board carrying mains voltage due to its orientation having been rotated 90 degrees with respect to the original layout.

 Since all of these AWGs still use class II smpsu boards inside of a plastic enclosure, they still conform to the requirements of a class II mains appliance, allowing it to be powered with a twin core mains lead absent a safety earth wire. In this case, the best way to make use of a C14 or C6 three pole mains connection upgrade is to wire a 'static drain' resistor from the PE tag to the zero volt rail using a lowish value resistor in the range 1K to 10KR to attenuate the half live mains touch / ESD hazard down to just 50mV to 500mV rms respectively on 240v mains supplies and assuming a worst case BNC shield interconnect resistance as high as one ohm (which condition also attenuates unwanted ingress of noise and random DC voltage offsets from the mains earth wiring by some 60 to 80dB).

 None of these Feeltech AWGs actually require a safety ground connection other than as a means of providing a safe and convenient connection for a 1 to 10 kR static drain / half live mains leakage shunt path to ground. And, of course this also applies to its many clones and copycats built into cheap plastic enclosures demanding no bigger a power requirement than what a 10W rated smpsu (whether internal or from a cheap wallwart) can supply.

 These typically employ a 1nF Y cap to shunt the switching noise that is capacitively coupled via the interwinding capacitance of the transformer onto the dc output back to its source in the mains rectified HT supply powering the switching circuit in order to satisfy EMC requirements. The use of direct rectification of the incoming mains supply creates a half mains voltage ripple with respect to the neutral and ground which this Y cap passes onto the DC output terminals as a high impedance voltage source.

 At 50 Hz the capacitor has a reactance of just over 3M ohms limiting the touch current to a maximum of 37μA on a 240v supply. Higher rated class II smpsus (20 to 30 watt range?) typically use a 2.2nF Y cap which has a reactance of just under 1.5M ohms resulting in a maximum leakage current in the region of 80μA on 240v supplies. ISTR a figure of 85μA being the maximum allowed for class II certified psus which limits the maximum power rating of a cheap 'n'cheerful class II smpsu using a transformer not equipped with an inter winding shield foil layer to screen the secondary winding(s) from the high energy switching noise on the primary.

 Anyway, for those interested in eliminating half live leakage/ESD issues or the manufacturers' misguided grounding measures to eliminate such issues which cause mysterious random DC offsets and unwanted ingress of hum and noise from the mains earth wiring, that's pretty much all you need to know to cure your bargain basement Chinese AWG of its maladies.

[McBryce]

Hi Johnny,
      you are aware that the Siglent SDG2042 isn't in any way based on the design of the Feeltech device. It also isn't in a plastic case, based on the same design or even follow any of the same design decisions? It's a €600 device that you are comparing to an €80 toy? I would expect that any issues Siglent need to remedy will be approached at a whole different level than the guys at Feeltech would.

McBryce.

[Johnny B Good]

Hi Johnny,
      you are aware that the Siglent SDG2042 isn't in any way based on the design of the Feeltech device. It also isn't in a plastic case, based on the same design or even follow any of the same design decisions? It's a €600 device that you are comparing to an €80 toy? I would expect that any issues Siglent need to remedy will be approached at a whole different level than the guys at Feeltech would.

McBryce.

 Damn right!  Oh, and BTW, thanks for giving me the opportunity to explain my (justifiably) jaundiced view of Siglent's AWG/function generator product line.

 Quite frankly imo, Siglent should be hanging their collective heads in utter shame over the shambles of a UI that's been imposed on their whole AWG range. I can't believe that the DDS technology is any less capable than that used in the Feeltech models from the FY6600 series onwards in regard of the 1μHz resolution across a 14 digit range (DC through to 60 or the 100MHz maximum of the latest FY6900-100M).

 It's a complete and utter mystery to me as to why they should have deemed a maximum of 11 digits via the keypad (with a mere 8 digits displayed maximum if using the rotary encoder) should be sufficient in a function generator that can be locked to an external atomic reference that can provide the required level of stability and accuracy to justify a 1μHz step size at least up to and around the 10MHz region of interest.

 It's not as if the display lacks room for the additional six digits required to allow 1μHz steps up to the 100MHz region to be displayed if they abbreviate "Frequency" to the industry standard form, "Freq". If these AWG models had lacked an external reference socket, these limitations wouldn't have been an issue.

 I can only guess that the external reference socket (unfortunately cursed with a b-directional function to allow it to export its internal +/-1ppm 10MHz clock by default to create mayhem on an existing linear bus section of a lab reference distribution system) had been a late addition in its development program wherein the additional 6 digits' worth of resolution had failed to make it into the original firmware design brief, leaving only the matter of this add-on feature as a last minute hack job to add an ill considered set of user options into the UI.

 As you can see, this is the other major fail in these Siglent AWGs. If they'd invested a few more cents into adding a separate reference output socket to eliminate this bi-directional clash scenario, there would not have been any such risk of polluting a lab reference distribution with such a low grade 10MHz source.

 I'm well aware of the benefit of synchronising to a single master clock even when it falls far short of the accuracy expected of a dedicated lab standard 10MHz reference but when you're cost cutting on sockets, you should at least put a little more thought into the management of your bi-directional reference socket's functionality than has been demonstrated here.

 I'm sure there are a few more shortcomings in the UI but these two are the most egregious imo considering it's now some five six! years and counting. I'll leave the lesser shortcomings to be reported by others for the time being.

 I've already had numerous email discussions just over 11 months ago with Thomas Rottach, head of Siglent's European sales division, involving an SDG1032X I'd just purchased from Telonic which had a faulty VCXO (it looked like a flicker noise ripple effect on its EFC pin) as a result of Telonic's own "technicians" and the supposed technicians they'd consulted with at Siglent EU failing to understand my very clear sequence of DSO screengrabs and video evidence of the problem. In the end, I gave up on getting a price adjusted exchange for an SDG2042X out of Telonic and insisted on a refund on the basis of it being unfit for purpose.

 To be fair, Telonic had been as much a victim of this muddle as I had and obliged my request with alacrity, even to the extent of collecting the ailing SDG1032X at their own expense. I'd have preferred to upgrade my spending with them but the process had dragged on for several weeks by then, so I ordered the SDS2042X from Labtronics from whom I'd purchased my SDS1202X-E two years earlier.

 Within a week of placing the order, I had a brand new SDG2042X in my hands. As per usual, I had picked the worst day of the week, possibly a bank holiday weekend, to place my order and expect a 3 day delivery schedule. Anyhow, the first thing I did was to repeat my "Frequency accuracy and stability tests" to verify that it was completely devoid of the 'jitter issue' I'd discovered in the SDG1032X following this up by confirming that this shared exactly the same UI defects I'd mentioned to Thomas Rottach (head of Siglent's European Sales Division) in no uncertain terms during my many email exchanges.

 He had told me that he would pass on my UI related concerns  to the Siglent factory technicians for consideration in the next firmware update. I've just checked for the latest firmware update but I see it's the same one that had been released almost two years ago which I'd downloaded back in November last year.  So much for Mr. Rottach's assurances.

 There did happen to be an external/internal reference clock related fix in this last update but it related to disabling the automatic reversion to its internal clock reference on detection of the loss of the external reference, with an icon to signify this state and a pop up message whenever you try to turn a channel on.

 This is far from the solution I'd hoped to see and strongly suggests that the hardware involved is incapable of using its internal reference without polluting the connection to the external reference network with a duplicate of its own clock. IOW, an issue that can't be properly fixed with a firmware update, hence the kludge solution used in this last (final?) update.

 Fortunately, this is not the end of the world since the problem can be worked around by giving it a dedicated feed from a 10MHz distribution amp that can provide the necessary isolation between its ports. It's the strangely crippled frequency entry UI that's the most egregious shortcoming of these function generators and that's a fix that does look amenable to a firmware update.

 Who knows! Perhaps one day soon, we may get to see the full frequency setting capabilities of Siglent's AWG range finally released from its prison of incompetent firmware coding to match the reach of that of the Feeltech FY6600 and its successors. Despite the lack of a numeric entry keypad (or possibly because of this lack) Feeltech's solution to providing the means to adjust the frequency in 1μHz steps all the way from DC to 60 or even 100 MHz looks like an act of genius compared to Siglent's rather lackadaisical effort.

 Yes, as supplied, the last 8 digits were a complete waste of space until it had been upgraded to an OCXO clock reference with an external 10MHz clock reference socket. It seems to me that Feeltech had designed the FY6600 to the highest standard and then proceeded to apply their cost cutting compromises only to those components that could be readily upgraded by the keen cash starved hobbyist (Feeltech's target demographic) to polish this rough diamond into something of a gem that could ultimately outshine the SDG1042X and its contemporaries. A numeric keypad is a nice luxury but, at the end of the day in this class of test gear, that's all it is, a non-essential 'luxury'.

 Don't get me wrong, I'd retire the FY6600 in a heartbeat as soon as Siglent fix their crippled frequency setting UI. Until that happy day arrives, the FY6600 will remain my 'go to' function generator and my SDG2042X can continue its usefulness as a plinth for the FY6600.

[nike75]

Hi,
now I have a totally bricked device.

Monts ago I found troubles with power supply on my SDG2042X. After long working and try to power cycle (power off and power on again) PS enter in starting loops, voltages is rised but not to nominal values and restart. Fan also on and of every second. If device is switched off, relax few minutes and power on again - all is OK.

Currently I found problem and fix. I will send separately post for this patch.

But many times before after these power loops, my device lose config values, even more - startup counter is zeroed few times.
For safety reasons I make backups on all mtd's. During nandump I found several ECC corected bitflips.

Today I decided to do some NAND test and after check all mtd's with nandtest, ls command go to return invalid argument, and SDG become unresponsibled via buttons.
These were the last moments of his life - after power cycle only fan working, no LCD, nothing.

All voltages are OK, CPU led light in green, 25MHz is OK

At the moment I'm not sure where to start 

[thinkfat]

"nandtest" does destructive tests. If you didn't use the "-k" option, your NAND content is now "gonsky".

[nike75]

Yes, I actually use nandtest without any option   

Now my nanddump files can help me?

[thinkfat]

Yes, I actually use nandtest without any option   

Now my nanddump files can help me?

I'm not super familiar with raw nand handling, potentially they might help, if you are still able to get to a console prompt with access to the dump files. If they're raw binaries, you might be able to restore the NAND devices using the "nandwrite" command or "dd" (whatever you have access to).

[MathWizard]

Ok why was I thinking this was safe to do ?? I wanted to test a transformer, to find the output voltages. So I put my sdg2042 on 20Vpp, on default HiZ , and hooked it directly to the transformer primary, without using any DC blocking, or a resistor to limit current.

I've done it a few times, it still seems to work fine, but what kind of current was I putting through the low winding resistance of the trans I don't want to operate it on the edge or put it into some hard current limiting.......is that what I was doing ? What current does the highZ output really do when near shorted?

[mawyatt]

If you are working where you need more current or voltage try using an external Amp like the Juntek DPA-1698 or DPA-2698. For even higher levels, we designed a couple external amps for HV outputs ~(+-90V) and Higher Currents, although we haven't built the High Current version yet (no immediate need although the PCBs have been designed and fabricated and parts are here), the HV version has been built and used.

https://www.eevblog.com/forum/testgear/awg-buffer-amp/msg3368650/#msg3368650

https://www.eevblog.com/forum/testgear/cheap-diy-high-powervoltage-awg-buffer-amp/msg3449784/#msg3449784

https://www.eevblog.com/forum/testgear/hv-buffer-amp-for-awg/msg3623953/#msg3623953



Best,

[bdunham7]

I've done it a few times, it still seems to work fine, but what kind of current was I putting through the low winding resistance of the trans I don't want to operate it on the edge or put it into some hard current limiting.......is that what I was doing ? What current does the highZ output really do when near shorted?

It doesn't matter what the output impedance 'setting' is, it just has a 50 ohm impedance.  So 20V_p-p through a 50 ohm resistor is what you get.

[2N3055]

I've done it a few times, it still seems to work fine, but what kind of current was I putting through the low winding resistance of the trans I don't want to operate it on the edge or put it into some hard current limiting.......is that what I was doing ? What current does the highZ output really do when near shorted?

It doesn't matter what the output impedance 'setting' is, it just has a 50 ohm impedance.  So 20V_p-p through a 50 ohm resistor is what you get.

This..

But thing to worry is not that, but fact that highly inductive load on the generator output can create kickback voltages and currents high enough to maybe damage output of the generator.. So depending on trafo used a snubber and/or clamp might be in order. Not to mention what happens when you get something in resonance and voltage starts climbing through the roof...

So safe practice would be to start with smaller amplitude and check with a scope how output from gen/input to the trafo looks like.

[tautech]

Yes, I actually use nandtest without any option   

Now my nanddump files can help me?

Check your PM's.

[bdunham7]

I apologize if this issue has been raised, and perhaps dealt with, but I could find no specific mention of it elsewhere.

The subject came up about 'frequency coherence', meaning if you selected, for example, 10Hz pulse on CH1 and 10MHz sine on CH2, would the CH2 output frequency be an exact 10⁶ multiple of CH1.  So I had to try it, of course.

The answer, to my surprise, was 'maybe'.  When I first selected the two and set up a scope triggering on the 10Hz pulse, they were coherent and had no drift.  Fiddling with the higher frequency, I was able to get it to show some drift at 10.000010 Hz at certain times, and then once at exactly 10MHz.  I can't reliably replicate those, so I'll move on for now.  What I eventually found is that if I pressed the MOD button, even with 0% AM modulation, the 10MHz signal would start to drift relative to the 10Hz signal, by about 1/3 cycles per second or 0.3ppm.  Hmmm.

I decided to check the frequencies against the clock and I discovered that in the normal startup configuration, a 10MHz selected output was ~0.3ppm low compared to the 10MHz clock.  The 10Hz pulse was also off by the same amount, so that the two outputs were synced, but the clock drifted relative to them or vice-versa.  An external clock made no difference.  However, when I press the MOD button, again with no modulation, the 10MHz CH2 output and the clock would then coincide perfectly, but would drift 0.3ppm relative to the pulse output. 

I recalibrated the SDG2042X clock to my OXCO in my counter and it only took a slight correction, less than 1ppm.  Th original DAC code was 18022 and I ended up with 17865.  I tried different DAC codes from 1 to 20,000 and while it changed the clock frequency, it made no difference to the observed behavior.  So this behaviour seems to be baked in.

Can anyone confirm that this is normal?  Any ideas why it is set up this way?

[nctnico]

I can give it a try somewhere in the next couple of days.

[bicycleguy]

...
I can't reliably replicate those, so I'll move on for now.  ...   Can anyone confirm that this is normal?

Normal for SDGxxxxxx for last ~ 6 years.  Stems from the lack of displaying the actual precision.  See Johny B Good summary below a few quotes.  Any time you touch the rotary knob the digits stored but not displayed are in doubt.  The only way to be sure is to punch in the frequency on the keypad to the desired resolution using the keypad.  I can't find my notes but I think 10,000,000.01 can be shown to be different than 10,000,000.02 but the display only shows 10,000,000 after you input.  This easily shows up on an oscilloscope comparing output to an OCXO.

Any ideas why it is set up this way?

To confound and frustrate the user to no end, especially when trying to debug a home grown GPSD OCXO.

[bdunham7]

Normal for SDGxxxxxx for last ~ 6 years.  Stems from the lack of displaying the actual precision.

Perhaps that happens in some cases, but in this case I can enter 1-0-MHz with the keypad and my clock and output will be off 0.3ppm, but if I press MOD, they align--but the slower 10Hz pulse that I got by pressing 1-0-Hz will not.

[Hexley]

I can reproduce this with an SDG1032X.: confirmed stable phase of 10 MHz sine and 10 Hz pulse with MOD turned off, then a steady phase shift between those two signals with MOD set to AM 0%.

Another interesting test: Change channel 2 to 100 Hz. Change the modulation from channel 1 to channel 2. Set the modulation to AM 50%, 1 Hz. Trigger the scope on channel 2, and observe two things: (1) the 10 MHz channel appears to jump from one phase offset to the next to the next, with 15 steps before it covers 360 degrees, and (2) those discrete phase offsets have a small 1 Hz phase modulation of their own. The attached photo shows both phenomena.

[nctnico]

I apologize if this issue has been raised, and perhaps dealt with, but I could find no specific mention of it elsewhere.

The subject came up about 'frequency coherence', meaning if you selected, for example, 10Hz pulse on CH1 and 10MHz sine on CH2, would the CH2 output frequency be an exact 10⁶ multiple of CH1.  So I had to try it, of course.

The answer, to my surprise, was 'maybe'.  When I first selected the two and set up a scope triggering on the 10Hz pulse, they were coherent and had no drift.  Fiddling with the higher frequency, I was able to get it to show some drift at 10.000010 Hz at certain times, and then once at exactly 10MHz.  I can't reliably replicate those, so I'll move on for now.  What I eventually found is that if I pressed the MOD button, even with 0% AM modulation, the 10MHz signal would start to drift relative to the 10Hz signal, by about 1/3 cycles per second or 0.3ppm.  Hmmm.

I decided to check the frequencies against the clock and I discovered that in the normal startup configuration, a 10MHz selected output was ~0.3ppm low compared to the 10MHz clock.  The 10Hz pulse was also off by the same amount, so that the two outputs were synced, but the clock drifted relative to them or vice-versa.  An external clock made no difference.  However, when I press the MOD button, again with no modulation, the 10MHz CH2 output and the clock would then coincide perfectly, but would drift 0.3ppm relative to the pulse output. 

I recalibrated the SDG2042X clock to my OXCO in my counter and it only took a slight correction, less than 1ppm.  Th original DAC code was 18022 and I ended up with 17865.  I tried different DAC codes from 1 to 20,000 and while it changed the clock frequency, it made no difference to the observed behavior.  So this behaviour seems to be baked in.

Can anyone confirm that this is normal?  Any ideas why it is set up this way?

Today I have done test at a customer who has a SDG2024X and a Cesium clock. I compared the Cesium clock output and a 10MHz output from the SDG2024X (which using the Cesium clock as a reference) using an oscilloscope but these are locked together just fine. A difference of .3ppm would show up as a continuously moving signal on an oscilloscope. I'm not sure what is happening exactly in your SDG2024X; maybe the amplitude you are feeding into the external input is too low, the OCXO has a problem or your SDG2024X is faulty. The SDG2024X I used runs firmware version 2.01.01.23R8

[bdunham7]

Today I have done test at a customer who has a SDG2024X and a Cesium clock. I compared the Cesium clock output and a 10MHz output from the SDG2024X (which using the Cesium clock as a reference) using an oscilloscope but these are locked together just fine. A difference of .3ppm would show up as a continuously moving signal on an oscilloscope. I'm not sure what is happening exactly in your SDG2024X; maybe the amplitude you are feeding into the external input is too low, the OCXO has a problem or your SDG2024X is faulty. The SDG2024X I used runs firmware version 2.01.01.23R8

Thanks for checking. So, Cs clock connected to clock input of SDG2042X, external clock selected, oscilloscope monitors the Cs clock signal and the 10MHz selected out and they are perfectly synced.   Hmmmmm.  I have version 2.01.01.35R3B2, which unlike the earlier version in yours will not revert to the internal clock if the external clock signal is inadequate.  And I still have the exact same issue with both the internal and external clock.  Maybe when I get a chance I'll roll back the firmware.

[Blue]

I run 2.01.01.35R3B3.
The older ones have a problem with Labview.

Ask Tautech for an update. Perhaps there is even a newer one.

[blurpy]

I'm going to build the Elenco AMFM108CK radio kit, and in the manual we can see several figures like this:



There is a capacitor between the function generator and the DUT.
So guess this is to block DC entering the generator when sending a signal into the middle of a circuit. Is this common, and a good way to protect the generator?

Full manual available here: https://www.elenco.com/wp-content/uploads/2017/10/AMFM-108CK_low-res-2.pdf