EEVblog Electronics Community Forum
Products => Test Equipment => Topic started by: Hardware on August 26, 2020, 04:45:43 pm
-
When I was still at school, I built an ELV function generator with frequency counter about 30 years ago. I was already dissatisfied with the result regarding signal quality or counting accuracy at that time and could not be compared at all with today's devices.
So now, even without a concrete project, I want to buy something that I do not have to be annoyed about and that is fun until retirement.
So, I am hesitating between buying a new Siglent SDG2042X and buying an Agilent 33220A on eBay.
The signal quality, low distortion, low soundfloor, frequency stability is important to me. I like the integrated frequency counter of the Siglent, but maybe a separate device is better.
I could not find a comparison of these devices in the net so far, but maybe someone has worked with both devices and can make a valid statement.
-
In ~1980 we needed to create some unique and interesting frequency domain waveforms and specialized test equipment for a new developed real time Fourier Transform device based upon the Chirp Z Transform. Since nothing we were aware of could do this, we developed what could be called an arbitrary waveform generator based upon a couple very expensive high speed 16 bit hybrid DACs and a massive array of fast SRAM, the inverse FFT was used to translate the arbitrary frequency domain waveform into a time sequence to produce the desired frequency domain waveforms.
I'm retired like you, and have a history way back of dealing with AWG as mentioned (we were also doing Microwave/RF sub-sampling back then). I wanted an AWG that wouldn't be too expensive but still maintain a high level of signal quality and custom waveform capability. After deciding on an DSO from Siglent (SDS2102X Plus, 2104 was backordered) and impressed with how well this instrument works (exceeding my expectations) I went ahead with the Siglent SDG2042X.
I'm also impressed by the SDG2042X, very clean and versatile waveforms. So having got 2 hits in a row from Siglent I just pulled the trigger on the SSA3021X Plus and also very impressed with this fine instrument :)
One of the first things I did with the SSA was check the two tone output from the AWG. Here's a screen shot of the SSA with the AWG using the "Wave Combine" function with Ch 1 at 10MHz -10dBm and Ch 2 at 10.0000020MHz at -10dBm, and another at 1MHz. Not outstanding performance but good enough for what I need.
The AWG has a nice feature with the ability to have the internal reference frequency (TCXO) adjusted by DAC commands, so you can adjust the ref externally. I used the SSA Ref Output as the reference and adjusted the AWG internal reference this way (of course you can synch them up as well). It also works well with the SDS2102X Plus with Bode Plots via USB, except they are very very slow!!
I wish the waveform generation software was available for the MAC, and the touch screen isn't that easy to use.
I'm picky when it comes to electronics, especially when paid for out of my retirement funds. The fact that I've purchased 3 different Siglent products in the past month speaks very highly of their instruments.
Anyway, I don't think you'll be disappointed with the SDG2042X.
Later I plan to "upgrade" the AWG and SSA to enable greater BW, I've already done the SDS2102X Plus and it wasn't hard but evidently the SDG2042 and SSA3021 are much more involved.
Best,
-
I have worked with the Agilent (though not extensively) and I have the Siglent at home. I haven't had any issues with either of them. The interface of the Siglent is heavily inspired by the newer Keysight generators. For more sophisticated configurations (sweeps, modulation, harmonics...) it is nice to have all the important settings visible at once. But the Agilent does the job.
I'm not sure how the specs compare in detail. For general purpose use they will likely both be fine. If you have specific requirements you will need to check the datasheets and see if they meet your needs.
There are some bigger differences that might impact your decision:
- the Siglent has two channels
- the Siglent can be hacked to do 120MHz sine (to SDG2122X)
- the Siglent has 8Mpts memory for AWG, the Agilent only has 64kpts; for periodic signals this may not matter much, but if you want to generate long transients (or replay something you captured with an oscilloscope) this is a big difference
- Keysight has a rather nice and polished software for creating arbitrary waveforms (but I think it costs extra); the old Siglent software was horrible, but I think they have a new one. If you use Python or MATLAB anyway you may not care.
- the Agilent has a floating output, the Siglent is ground-referenced; depending on what you want to do, this may be a very big deal
Both will be a huge step up from the old analog ones in terms of frequency stability, amplitude accuracy and distortion.
-
Dear mawyatt,
Many thanks for your quick, detailed explanations and of course the photos from the spectrum analyzer. They look really good.
Comparative pictures of the Agilent would of course be great.
But of course you can only show what you have. Nevertheless, thank you again!
For the Agilent products, the documentation is available online and I think it is excellent. It contains a circuit diagram and a parts list of the components used. Accordingly, this should at least be a good prerequisite for repairing the device. This point is probably also important to consider with an older eBay device.
I haven't seen any circuit diagrams for Siglent products. But maybe this is also due to the fact that you don't actually want to repair any devices or can only rarely repair them.
It may be the same with current Keysight devices, at least with regard to the circuit diagrams with parts lists.
Perhaps there will be a few more statements about the Agilent model.
Best wishes
-
I haven't seen any circuit diagrams for Siglent products.
You won't other than users reverse engineered schematics.
We can't get anything other than service manuals that are available freely online.
Replacement PCB's are available should they be required.
http://siglentna.com/wp-content/uploads/dlm_uploads/2017/10/SDG2000X_ServiceManual_SM0202X-E01A.pdf (http://siglentna.com/wp-content/uploads/dlm_uploads/2017/10/SDG2000X_ServiceManual_SM0202X-E01A.pdf)
But maybe this is also due to the fact that you don't actually want to repair any devices or can only rarely repair them.
It's not often SDG2000X models fail.....we've never had one in NZ but I believe there are one or 2 in the SDG2042X thread that have been damaged by excessive reverse voltages beyond the RV protection limitations. When this occurs it's normally restricted to the output stage where the devices used should be outlined somewhere in the SDG2042X thread:
https://www.eevblog.com/forum/testgear/the-siglent-sdg2042x-thread/ (https://www.eevblog.com/forum/testgear/the-siglent-sdg2042x-thread/)
-
I can echo with other Siglent owners that the SDG is a decent AWG that is good enough for most people.
However, if sharp square waves and pulses are important, even with the EasyPulse, its 9 ns rise time and 150 ps jitter may not be good enough for some. There is where the higher priced Keysight Tureform 33600A series shines, its 2.9 ns rise time and 0.5 ps jitter makes a difference, but you are now looking at something where you pay more sales tax than what the Siglent costs.
Not knowing your use case and exact budget, that is something you should be aware of. The reason I bring up the Keysight is because you mentioned "The signal quality, low distortion, low soundfloor, frequency stability is important to me." without quantifying what your criteria is. The Siglent may not be the right solution depending on your needs.
The 33220A is a legacy design dated back to the Agilent days, there is nothing special about its spec, so I think the Siglent should compare well against it.
Given you do not have a concrete project yet, I agree with others that you can buy a Siglent for now, and upgrade to higher end gears if the need arises.
Hope this helps.
-
I can echo with other Siglent owners that the SDG is a decent AWG that is good enough for most people.
However, if sharp square waves and pulses are important, even with the EasyPulse, its 9 ns rise time and 150 ps jitter may not be good enough for some. There is where the higher priced Keysight Tureform 33600A series shines, its 2.9 ns rise time and 0.5 ps jitter makes a difference, but you are now looking at something where you pay more sales tax than what the Siglent costs.
At that point the SDG6022X stands up but at a totally different price point than SDG2042X. :o
-
Tautech has a very good point, the 6K series may be a good compromise to get better spec but still keeping the price very reasonable. If I were to rebuy another Siglent AWG, that's what I would do to get sharper square waves and pulses. You don't get the crazy low jitter of the Keysight, but if it is not important to you, it is a very good option.
Tautech, can the jitter spec be improved if you drive it with an external GPSDO or Rubidium reference? As I don't see an OCXO being an option for the 6K series, which would have been a great additional.
-
Tautech, can the jitter spec be improved if you drive it with an external GPSDO or Rubidium reference? As I don't see an OCXO being an option for the 6K series, which would have been a great additional.
Dunno, 100ps jitter is getting down where it's getting not straightforward to measure.
Maybe someone has posted about using Ext ref to reduce jitter in the SDG6000X thread.
https://www.eevblog.com/forum/testgear/siglent-sdg6000-series-awg_s/ (https://www.eevblog.com/forum/testgear/siglent-sdg6000-series-awg_s/)
At this time only the SSG5000X RF gens offer internal frequency ref options.
-
Not sure if this is good advice but:
https://www.microsemi.com/blog/2016/08/24/how-to-measure-jitter/#:~:text=The%20standard%20procedure%20for%20measuring,peak%2Dto%2Dpeak%20values. (https://www.microsemi.com/blog/2016/08/24/how-to-measure-jitter/#:~:text=The%20standard%20procedure%20for%20measuring,peak%2Dto%2Dpeak%20values.)
"The standard procedure for measuring period jitter involves randomly measuring the duration of one clock period 10,000 times, and using the recorded data to calculate the mean, standard deviation and peak-to-peak values."
I'm not convinced I have the data or skill to do fully follow the above procedure but fwiw here are some images showing a generator outputting a signal at 9 MHz into a frequency counter. Both are locked to a GPSDO 10 MHz reference signal. There might be some magic in the generator but I'm pretty some of the magic comes from the GPSDO (which is currently receiving a pretty weak signal due to a tree).
I don't have a good way to measure and record the duration of each of 10,000 random clock periods but here are the cumulative results of 14,555 sequential samples with 10 second gates. While this doesn't speak specifically to jitter the numbers indicate that over 145,550 seconds (about 40 hours) the 9 MHz signal is on average off by about 10 uHz. Unless I'm missing something (always fairly likely) the ability to measure stuff to a high degree of resolution (and I think also with a with relatively high degree of accuracy) with relatively affordable test equipment is pretty amazing. I realize the generator is somewhat expensive but I have a hunch that using a Siglent 2042X (less than $500) with the same GPSDO and frequency counter (each of which are less than $200) would likely yield similar results. Might be interesting to see such a test with a 2042X.
-
It also works well with the SDS2102X Plus with Bode Plots via USB, except they are very very slow!!
Perhaps it is good to discuss and look this bit more deeper in some day and in some other thread what handle SiglentBodePlot II features and performance and even independent of Siglent scope model.
Of course Siglent can do much faster simple and low dynamic range BodePlot what some other manufacturer have done just like for checkbox feature for shiny sales ad's what Keysight launches - just bit before Siglent have get ready its own MUCH more intelligent BP I and soon later lot of higher performance BP II.
Also it have two different working mode where speed is different. Autolevel and fixed level. Fixed level is naturally lot of faster but with less total dynamic range due to receiver ALC off.
It is good to remember it do not BP just like taking scope xt, hide it, and and draw peak shape. Siglent BP is more complex but also performance walk easy over most competitors in low price segment, look example Keysight low price scopes BP joke or some GoodWill and today many others, even crap hantek. First it is up to three channel frequency selective sweeping receiver based with proportionally adjusted RBW and where automatic dynamic range is far over 100dB inside one sweep. There is many things what take some time because all is made in Oscilloscope hardware, not in dedicated SFRA hardware.
If reason, lets continue in some other thread for this.
End of OT. ;)
-
Jitter.
I remember rf-loop showing examples of simple jitter measurement and a little search found his examples.
Of course this first one shows the sum of all jitter for the DUT and the measurement system.
(https://www.eevblog.com/forum/testgear/feeltech-fy6600-60mhz-2-ch-vco-function-arbitrary-waveform-signal-generator/?action=dlattach;attach=453376;image)
Similar but examining 1% phase shift setting jitter between both AWG channels with infinite persistence:
(https://www.eevblog.com/forum/testgear/the-_official_-siglent-sds1202x-e-thread/?action=dlattach;attach=334267)
Respective threads:
https://www.eevblog.com/forum/testgear/feeltech-fy6600-60mhz-2-ch-vco-function-arbitrary-waveform-signal-generator/msg1600870/#msg1600870 (https://www.eevblog.com/forum/testgear/feeltech-fy6600-60mhz-2-ch-vco-function-arbitrary-waveform-signal-generator/msg1600870/#msg1600870)
https://www.eevblog.com/forum/testgear/the-_official_-siglent-sds1202x-e-thread/msg1262719/#msg1262719 (https://www.eevblog.com/forum/testgear/the-_official_-siglent-sds1202x-e-thread/msg1262719/#msg1262719)
-
I can only thank you very much for the many comments.
Concerning the SDG2042X I could not find any statement about the operating noise and the short circuit resistance despite a detailed study of the documents.
Can anyone contribute to the clarification?
Otherwise I could only find the Total Harmonic Distortion after comparing the given specifications. The Agilent with 0.04% is better than the Siglent with 0.075%.
-
Concerning the SDG2042X I could not find any statement about the operating noise and the short circuit resistance despite a detailed study of the documents.
Why would this matter when these are short circuit protected ? :-//
IMHO reverse voltage protection and its max limits are more important for an AWG.
-
If the output is short circuit proof everything is fine. I assumed it, but I just could not find the statement anywhere.
How high is the reverse voltage protection?
Can you also make a statement about the operating noise?
The device likes to run for a few hours.
-
If the output is short circuit proof everything is fine. I assumed it, but I just could not find the statement anywhere.
How high is the reverse voltage protection?
I can't quickly find that spec other than this description from the user manual:
When overvoltage protection occurs, a message will be displayed and the output is disabled.
The absolute value of input voltage is higher than 11V±0.5V when the amplitude of the generator is higher than or equal to 2Vpp or the DC offset is higher than or equal to |3VDC|.
The absolute value of input voltage is higher than 4V±0.5V when the amplitude of the generator is lower than 2Vpp or the DC offset is lower than |3VDC|.
How Overvoltage Protection works is described in the user manual on P126
https://www.siglenteu.com/wp-content/uploads/dlm_uploads/2017/10/SDG2000X_UserManual_UM0202X-E02C.pdf (https://www.siglenteu.com/wp-content/uploads/dlm_uploads/2017/10/SDG2000X_UserManual_UM0202X-E02C.pdf)
Can you also make a statement about the operating noise?
The device likes to run for a few hours.
Not silent but not offensive.
Unless there is more background noise you will know when it's ON. Much like the noise from my mini tower PC on the bench next to me.
-
The THD of my SDG2042X is quite a bit better than specified. My guess is that they decided against a tighter specification because then a proper audio analyzer would be required for performance verification.
0dbm into 50ohms, up to 9th harmonic
20Hz 0.020%
50Hz 0.009%
100Hz 0.007%
1kHz 0.006%
This is right at the limit of what my Keithley 2015 can measure, so take it with a grain of salt. So actually quite usable for many audio applications. A good soundcard will still be better though.
The datasheet specifies a maximum output current of +/-200mA, so with a 50 ohm output impedance and 10Vp maximum output voltage it will drive a short-circuit just fine without additional protection. Like most other AWGs it is not really made for driving low impedance loads though, the signal will be heavily attenuated.
Compared to some other test equipment I have, the Siglent is relatively quiet, but definitely audible. Mostly the airflow, no high-pitched fan noise. The Agilent is probably similar. I only used it in a lab environment which is not exactly quiet. But at least it didn't stand out.
-
Like most other AWGs it is not really made for driving low impedance loads though, the signal will be heavily attenuated.
Example if you have selected HiZ and 6V or 50ohm and 3V.
Yes and it follow just Ohms law. So it is also for low imp Loads but heavily attenuated, not unknown amount, but just nearly perfectly predictable.
Internal output impedance is Always 50 ohm. Independent of what external Load value you set in SDG menu.
So if your Load is 10 ohm. After "0 impedance" source generator it goes internally through 50 ohm resistor (yes this is there on board after end amp), total is now 60ohm, If internal 0 impedance generator is set for 6V output (Example if you have selected HiZ and 6V or 50ohm and 3V), then over your 10 ohm Load is 1V.
If you set your SDG for HiZ impedance then you set just this zero impedance generator voltage and over your HiZ load is same voltage. If you now connect external 50 ohm load, total is 100ohm and over your external load is half and over internal resistor is other half. Very simply. Heavily attenuated... better say, attenuated just like Ohms law say, no more, no less.
And yes there is
Im not perfectly sure if in SDG2kX can set also under 50 ohm load value in newest FWs. Load set value means only that you tell to SDG what load value you connect to output so that it can Display righ voltage over your Load as long as you have told truth to SDG about Load value. But least it can free adjust between 50ohm to 100k and Hiz.
Fact to remember. In ALL cases internal source impedance behind BNC connector is 50 ohm.
(Gout) ---> [ 50R ] ----> Chx BNC out
-
Indeed, more gernerally the voltage seen by a load with impedance \$Z_L\$ will be
$$V_L = \frac{Z_L}{50 \Omega + Z_L} V_0$$
where \$V_0\$ is the voltage shown with output set to Hi-Z (just a voltage divider).
You are right, the load impedance can be set between 50ohms and 100k (and Hi-Z), but not below 50ohms. In fact, it is slightly non-intuitive, because the menu only gives you the options Hi-Z and 50ohms and you have to press 50ohms and only then can you enter a different value (or if you click on the current setting on the touch screen).
Anyway, there should be no technical reason why it shouldn't go lower. So seems like a bug? Maybe the UI can't handle the low voltage levels that would result (it shows only one decimal place for mV values) so they chose to limit it.
It is only a convenience feature (you can always do the calculation yourself), but would still be nice if this could be fixed (@tautech).
-
Dear Mawyatt,
thank you again for your good comments. Would you please set your Siglent SDG2042X to Square 5Mhz, 10Vpp and upload a screenshot of your oscilloscope?
As you can see, I found someone who has an Agilent and sent me this.
I would expect the siglent to produce the signal better.
I am quite convinced of the siglent after all the comments, but I still have some doubts.
What do you think, will the signal quality be improved by increasing the bandwidth of the SDG2042X for square wave signals? This is only a license, i.e. software, and no better hardware.
Many greetings
-
What do you think, will the signal quality be improved by increasing the bandwidth of the SDG2042X for square wave signals? This is only a license, i.e. software, and no better hardware.
SDG2042X and SDG2120X Sguare and Pulse specifications are exactly same.
I believe* there is only one hardware and it is SDG2120X.
In HP example image something is very wrong.
People who have tested it do he really know what he is doing. Or is his HP generator broken.
This amount of ringing after rise and fall edges do not at all look ok. Perhaps he have forget impedance matching or some other things are out of order example in his "analyzer"... I do not know this analyzer he have used as oscilloscope.
33220A Square
Frequency range 1 μHz to 20 MHz
Rise/Fall time < 13 ns
Overshoot < 2%
Variable duty cycle 20% to 80% (to 10 MHz) 40% to 60% (to 20 MHz)
Asymmetry (@ 50% duty) 1% of period + 5 ns
Jitter (RMS) 1 ns + 100 ppm of period
SDG2000X Square
Frequency 1μ 25M Hz
Rise/fall times 9 ns 10% ~ 90%,
Overshoot 3 % 100 kHz,
Duty cycle 0.001 99.999 % Limited by frequency setting
Jitter (rms), Cycle to cycle 150 ps
Now if example HP period is 10us (100kHz signal)
there is 1ns +100ppm and jitter is now 2ns and this is RMS!
When we use pulse is it important what is RMS error in time. Least I need live with peak to peak max timing errors. Who is intersted how much one can be RMS late, I look watch and sayh how many time units something is just now late. If I am in this year around 5 second RMS before train start why sometimes train go without me and I am there looking train in horizon and I am not there. One my circuit gate is not interested if clock have 1ns jitter but if it is even once 7ns too late whole system crash. I can flush toilet these RMS specs. But manufactures sales supporting engineers take this RMS like fashion to decorate cake. There is places for RMS example when I am heating my room...
... with 10kHz signal jitter is 11ns RMS
and with 1s period jitter is as much as 100001ns RMS. Nearly like ancient 10 years old Siglent SDG1000 what produce square from internal sine using comparator. Is this HP gen using this method??
Also this HP gen have 50MSa/s clock.
Even when it have been and is still somehow nice generator but we are now 2020.
-
Why don't you download datasheets for both generators and compare them, spec by spec?
Also you questions are quite undefined sometimes. What do you mean by statement "better squarewave"?
For instance SDG2000X series has Rise/fall times 9ns for squarewave and 8.4n-22.4 s (adjustable) for pulses..
Agilent 33220A has 13ns for Squarewave and Variable edge time < 13 ns to 100 ns for pulses that is seriously less capability.
And if you plan to use 25 MHz squarewave, make sure you understand that 10ns edges will be clearly visible on signal with 40ns period. That squarewave won't be very square. And on Agilent with 13ns edges it will look even worse.. And it is not because signal quality is bad. It is just how it is.
Except for Agilent 33220A having isolated ground output, everything else is better on Siglent SDG2042X.
Also you get new, calibrated instrument, with good amplitude flatness, two channels and more AWG memory.
There are many things you can do with two correlated channels, also, that make one channel gen much less capable...
Unless you can get Agilent 33220A in a pristine condition and with all options enabled for less money than Siglent SDG2042X, to me there is no contest here. You should also take a look at Rigol offering in that range.
-
Would you please set your Siglent SDG2042X to Square 5Mhz, 10Vpp and upload a screenshot of your oscilloscope?
10Vp-p is the SDG2042X maximum into a 50 ohm load which it must be for the output to be properly terminated otherwise the ringing that is evident in the screenshot of the Agilent will result.
There is more to replicating a square wave on a scope than just connecting a square wave source. Everything must be correct.
Examples here that don't unfortunately include a SDG2042X that one of us needs to fix:
https://www.eevblog.com/forum/testgear/show-us-your-square-wave/ (https://www.eevblog.com/forum/testgear/show-us-your-square-wave/)
Maybe I can find time to post one there later @ 5 MHz, 10Vp-p.......but it will be with a 50 ohm external termination so not to damage the 50 ohm input of the scope.
Later....as attached.
-
Would you please set your Siglent SDG2042X to Square 5Mhz, 10Vpp and upload a screenshot of your oscilloscope?
10Vp-p is the SDG2042X maximum into a 50 ohm load which it must be for the output to be properly terminated otherwise the ringing that is evident in the screenshot of the Agilent will result.
There is more to replicating a square wave on a scope than just connecting a square wave source. Everything must be correct.
Examples here that don't unfortunately include a SDG2042X that one of us needs to fix:
https://www.eevblog.com/forum/testgear/show-us-your-square-wave/ (https://www.eevblog.com/forum/testgear/show-us-your-square-wave/)
Maybe I can find time to post one there later @ 5 MHz, 10Vp-p.......but it will be with a 50 ohm external termination so not to damage the 50 ohm input of the scope.
Is it max 5Vpeak in scope. No. It is 5V dc or RMS. There is just simple resistor and nothing else.
10Vp-p square you can safely input to scope. If it break then Siglent is responsible. There is not even time limit for this value. Siglent: 50 Ω ≤5 Vrms, ±10 V Peak.
This is just how much it is heating... 0.5W and it can be continuous because not told any time limit. So there is also lot of marginal for short time exceed before fail. ;) But still do not exceed except if you are evaluating Abs max/damage levels...
-
Thank you very much for the many comments.
There was definitely something wrong yesterday.
Here the corrected measurement.
-
Thank you very much for the many comments.
There was definitely something wrong yesterday.
Here the corrected measurement.
But still perhaps wrong. If this what I read there is what I believe...
Btw, it is also more nice if you use bit more fast timescale for see rising and falling edges speed.
But what is wrong here...image is attached in next message. (even when it do not affect so much in this but in theory, it is mistake what is fatal class mistake if anyone do it in real work in lab. Not affect very much for shape but it is still wrong and not only thing, it also affect of course level. )
Do you remember what level you set in generator, do this scope now show this level... why
Kindly
;)
-
Here is friend @Tautech image from other thread. There is external 50ohm in use because he perhaps afraid this signal level is risk for internal 50ohm. What it is not in real. But careful is careful and - it do not cost anything...
5MHz sqr SDG2kX
(https://www.eevblog.com/forum/testgear/show-us-your-square-wave/?action=dlattach;attach=1057272;image)
and btw this is detail from your last test with HP ;)
(https://www.eevblog.com/forum/testgear/siglent-sdg2042x-vs-agilent-33220a/?action=dlattach;attach=1057320;image)
Your HP scope already have internal 50ohm inputs available also but I do not know its acceptable level limits..
-
Dear Mawyatt,
thank you again for your good comments. Would you please set your Siglent SDG2042X to Square 5Mhz, 10Vpp and upload a screenshot of your oscilloscope?
As you can see, I found someone who has an Agilent and sent me this.
I would expect the siglent to produce the signal better.
I am quite convinced of the siglent after all the comments, but I still have some doubts.
What do you think, will the signal quality be improved by increasing the bandwidth of the SDG2042X for square wave signals? This is only a license, i.e. software, and no better hardware.
Many greetings
I'm not going to use 10VPP for reasons mentioned, so using 2VPP. SDS2102X Plus input terminated at 50 ohms, and SDG2042X set at 50 ohms. Also included some photos of my Tek 2465 analog scope for reference with input terminated at 50 ohms.
I realize that 10VPP across 50 ohms is probably OK on the DSO, I know it is OK on the Tek 2465 since I've repaired the input hybrid and the thick film termination resistor is very large and on an alumina substrate (actually part of the substrate) and conservatively rated at 1/2 watt, the whole input hybrid is on an alumina substrate. But not knowing the size and details of this termination resistor on the DSO I'm cautioning on the safe side.
Specs are specs, some are conservative, some absolute limits and others are lies :o
If a spec limit is exceeded or approached (like the voltage rating on a electrolytic cap) and the possible result is degradation, or reduced life, or affects another parameter negatively then I tend to be conservative unless I absolutely know the device or instrument (like the Tek 2465 input termination) details. I don't know how the new to me SDS2102X Plus is terminated, and what components used, so staying conservative for the moment.
Anyway, I didn't have a BNC 50 ohm termination, so just went with a lower VPP.
Best,
-
and btw this is detail from your last test with HP
The peak-to-peak voltage measurement gives it away as well. Apart from that it probably doesn't matter much. 50 ohms into 75 ohms is still 1:1.5 VSWR. I'm still not sure what happened in the first measurement. With a rise time of 13ns and a relatively short cable, reasonably well matched at least on the output side I wouldn't expect anything nearly as bad. The amplitude is way off as well.
Anyway, I am not sure this is really all that helpful. The rise/fall times and overshoot are specified in both datasheets, so you basically know what to expect. If you need significantly faster square waves, you will likely be looking for a dedicated pulse generator. The jitter on the Agilent is really not that great. On that note, don't expect wonders from the Arb function either, I believe it is basically a naive DDS implementation. So no fancy resampling ("Trueform") like on the newer Keysight (or Siglent) generators.
I think there is a good case to be made for older, used HP/Agilent instruments in many cases (I have quite a few myself). The 34401A e.g. lacks a few bells and whistles, but spec-wise it is still a respectable multimeter and you can usually get one for much less than even a new B-brand one. An good power supply will still be a good power supply. A new RF power meter from pretty much any brand will probably be more than 10x what my old Agilent one cost. For generators and oscilloscopes things are a bit different. There are a lot of cheap, high-quality DACs and ADCs available now and signal processing is essentially free. This has changed the boundary conditions quite a bit. If you can get one for a really good price, it is still a solid choice. If not, the only thing it still has going for itself is the floating output.
There is a lot to be said for the new Keysight generators. They have all the advantages of the Siglent, (two) floating outputs, nice polished software, significantly improved specs and for business use I'd rather rely on Keysight customer service. BUT if those things are not important to you, you pay a lot of money for something you don't need.
-
and btw this is detail from your last test with HP
The peak-to-peak voltage measurement gives it away as well. Apart from that it probably doesn't matter much. 50 ohms into 75 ohms is still 1:1.5 VSWR. I'm still not sure what happened in the first measurement. With a rise time of 13ns and a relatively short cable, reasonably well matched at least on the output side I wouldn't expect anything nearly as bad. The amplitude is way off as well.
Anyway, I am not sure this is really all that helpful. The rise/fall times and overshoot are specified in both datasheets, so you basically know what to expect. If you need significantly faster square waves, you will likely be looking for a dedicated pulse generator. The jitter on the Agilent is really not that great. On that note, don't expect wonders from the Arb function either, I believe it is basically a naive DDS implementation. So no fancy resampling ("Trueform") like on the newer Keysight (or Siglent) generators.
I think there is a good case to be made for older, used HP/Agilent instruments in many cases (I have quite a few myself). The 34401A e.g. lacks a few bells and whistles, but spec-wise it is still a respectable multimeter and you can usually get one for much less than even a new B-brand one. An good power supply will still be a good power supply. A new RF power meter from pretty much any brand will probably be more than 10x what my old Agilent one cost. For generators and oscilloscopes things are a bit different. There are a lot of cheap, high-quality DACs and ADCs available now and signal processing is essentially free. This has changed the boundary conditions quite a bit. If you can get one for a really good price, it is still a solid choice. If not, the only thing it still has going for itself is the floating output.
There is a lot to be said for the new Keysight generators. They have all the advantages of the Siglent, (two) floating outputs, nice polished software, significantly improved specs and for business use I'd rather rely on Keysight customer service. BUT if those things are not important to you, you pay a lot of money for something you don't need.
Agree the generators and scopes have benefited from the advances in chips, probably more than any other instrument. DACs, ADCs and almost unlimited computing power have made our new DSOs really outstanding, doing many more functions than just a scope. I haven't had the pleasure of using these DSOs in the past few years because while still working (retired now) I was teaching others and developing advanced chips sitting in front of a computer running Cadence.
After retiring I got the urge to get back into the lab and got a couple Tek 2465 scopes and a couple HP34401A DVMs to repair, all are working now. This created the desire for a DSO, then a AWG, then a SA. I have been impressed with the DSO (SDS2102X Plus), same with AWG (SDG2042X) and SA (SSA3021X Plus), all benefiting from advanced chips which drive the cost down but still maintain a high level of performance.
Also regarding the 34401A, I don't think the high end DVMs have benefited as much from chip advancements, sure the handhelds have but not high performance bench tops. The precision voltage references and resistors are still pretty much the same as the older instruments, and the architectures basically the same which make up the DVM "core". Yes you have nice displays and such, but a basic DVM function is to accurately measure a voltage, current or resistance, like a scope function is to produce an accurate repsentation of the applied waveform, and a generator is to produce an accurate waveform.
The very fact that my old 34401As have maintained very good accuracy compared to the new KS34465A is a testament to the design and construction of those old DVMs. I also believe that the new KS34465 has some of the old "DNA" from the 34401A and why I decided on this DVM as my baseline in-house standard which I'll use to cal the older two 34401A. Even designed and built up some precision references to assist in checking calibration between instruments, see photos.
While on this subject for the OP, I did compare the SDG2042X output with the KS34465A in Sine, Square, Triangle and DC waveforms, lets just say I was quite impressed :)
Did the same with SSA3021X Plus, same result. ;D
The SDG2042X is quite a AWG IMO, would I purchase it again with what I've learned....absolutely ;)
Best,
-
I can only thank everyone very much for the many really good comments.
I asked an Agilent supplier for the signal curves of the 5Mhz rectangle at 10Vpp to challenge the function generator a bit and maybe see round corners in the signal.
Since I did not have much to do with electronics, measurement technology and the questions in this context since my studies, I was not aware of the problems resulting from the high signal voltage.
That's the way it is, if you only deal with software, operating systems, configurations and malware every day for more than 30 years.
I prepare myself for my retirement, so that later no boredom arises.
This is only for explanation!
Now since yesterday I have the Siglent sdg2042X, HW-version 02-02-00-40-00, SW-version 2.02.01.35R3B2.
First I put my old, please don't laugh, Hameg HM412-5 back into operation and connected it to the Siglent (output HIZ).
Sinus, Dreick looks great. I was a bit disappointed with the rectangle.
I had expected it to become round in the corners at higher frequency. It does. But you can already see slight rounding at 1kHz. At 10kHz the rounding is already noticeable and gets worse with increasing frequency.
Of course my Hameg only has a bandwidth up to 20MHz but up to 100kHz it should be no problem.
The Hameg internal 940Hz tuning signal looks slightly better for my feeling.
Do you have an idea what could be the cause? Or is my demand too high?
Or am I missing another setting on the siglent. Besides frequency, amplitude and load I have not yet made any freeI have not yet made any settings except frequency, amplitude and load.
If True Arb would help I have not understood yet.
In the first picture you can see the Hameg at the top and the siglent signal at the bottom.
-
First you would need to terminate that coax on scope side with 50 Ohm pass-trough terminator, to make sure there is no reflections on cable. And then let's calculate: SDG2000X has 8 ns risetime, and your 20 MHz scope will have 16-19 ns risetime... And that if it's properly adjusted. You need roughly 100 MHz scope or better to verify risetime of that pulse with any accuracy.....
-
You need to terminate the cable at the scope end with a 50 ohm BNC thru, then set the AWG to 50 ohms output. The SDG2042X is capable of producing <9ns rise times as I've shown with a good cable and proper termination.
This is a really nice AWG and should serve you well. Next you'll want a nice DSO and you can follow all the threads on this, I settled on the SDS2102X Plus (wanted a 2104X Plus but they were and still are backordered). Then the list continues to a SA, VNA, Linear PS and so on ::)
Best,
Edit, I see 2N3055 posted a couple minutes before me, so yes follow exactly what 2N3055 says :)
-
I had tried a tee with a 50Ohm termination that I still had from the time when Ethernet was running over coax.
This did not cause any significant change.
Where should I buy the path through and where not?
-
Just to add to what has been said: the bandwidth of a square wave depends primarily on its rise time, not on its frequency. So your 100kHz square wave will have significant frequency content well beyond 20MHz, because it has 8ns rise time.
This is also the reason your calibration output looks "better". You can't see the fast edge on the SDG output, because the scope is too slow, but what you can see are the reflections from the unterminated cable (the little bump). The edges of the calibration output will be slower, so cable reflections don't really matter yet.
To see this, you can try to make the edges of the SDG slower: switch to pulse mode, set frequency as before, 50% duty cycle to get a square wave again. Now you can actually change the rise/fall times. Make them larger and see the bumps go away and at some point you will see the edge getting flatter on the scope as well as you go above the oscilloscope's own rise time.
EDIT: There is also a chance that the oscilloscope itself is adding some distortion. In theory it should have a Gaussian response with no overshoot. But part values drift over time. There are probably a whole bunch of trimmers in the input stage for frequency compensation.
-
I had a new Siglent sds2104X Plus delivered to me for testing this weekend. So I can do some tests.
Yes, the rise time is crucial for the required bandwidth of the oscilloscope.
As suggested by "switchabl" I changed the SDG to pulse mode and reduced the rise/fall time significantly (1us). Then the overshoots disappeared from the oscilloscope.
Many thanks to Switchabl for the tip.
By the way, I was able to switch between the termination with 50 Ohm BNC thru (Rohde&Schwarz/Hameg HZ22 and China P57) and my previous termination with a T-piece and 50 Ohm on one side
no change in the shape and size of the overshoots.
I can see the first small overshoot already at a rectangle 2Vpp from the SGS2042X (yellow) of 1kHz. But it is like a small line and too small to see it clearly on a photo. From 10kHz on it becomes clearer and at 100kHz up to 1MHz it is nearly unchanged concerning the shape. From 5MHz on the horizontal ones get some inclination and the overshoots are nearly gone.
I put the internal AWG with the same signal via cable to channel 2 (pink). The internal AWG doesn't produce the rise time of course but also doesn't produce the overshoots.
Can someone please send me the 1MHz from a SGS2042X?
I just want to make sure that the overshoots from the function generator are normal and I can go back to choosing a current oscilloscope.
During my first attempts with the new Siglent Digital Oscilloscope, I already stumbled over some software bugs.
Best wishes
-
1Mhz/2ns from SDG6052 on SDS2504X+ , 1Vpp @ 50 \$\Omega\$, 30cm BNC cable:
[attachimg=1]
-
Hello Mawyat,
I would be very happy if you would compare my last oscilloscope pictures with your results.
It should not give worse signals than yours, otherwise I can exchange it.
Good wishes
-
Hello Mawyat,
I would be very happy if you would compare my last oscilloscope pictures with your results.
It should not give worse signals than yours, otherwise I can exchange it.
Good wishes
Here you go, taken with SDG2042X and SDS2102X Plus, sorry don't have a nice SDS2104X Plus :-\
Trying to duplicate what measurements were on your screen.
Best,
-
Hello Mawyat,
thank you very much for the perfect pictures. I still find the operation of the oscilloscope a bit difficult.
With you the internal AWG was on C1 and the SGS on C2. With me it was exactly the other way round, but you can see that.
Basically your rectangle looks a lot better than mine at all frequencies.
You have much less overshoot.
But what I'm surprised about is how similar the two signals are on your side. Is it possible that you show the two outputs of the SGS2042X here?
As I said, I had the oscilloscope internal generator on the second channel. The internal generator is a bit slower at rise/fall times. And you don't see this difference.
I have found out that the overshoot depends on the output voltage of the SGS. At 1Vpp it is much lower than at 2Vpp.
My BNC cable is about 18 inches long. So this should not be the reason for the overshoot.
Kind regards
-
I took these with the SDG2042X AWG using Channel 1 and 2. I did not use the internal SDS2102X Plus AWG. Here's using the internal AWG (Pink) and SDG2042X (Yellow).
Best,
-
I am now sure to have found the reason for the overshoot probematics.
So it's not the cable or the termination, it's simply the signal amplitude and comes from the SGS2042X.
So if you slowly reduce the amplitude of 2Vpp in steps of 0.1V you suddenly hear a switching noise in the SGS and the overshoot is significantly reduced.
The oscilloscope is able to measure the overshoot (ROV in %).
Siglent gives a value of <3% in the specifications regarding the overshoot.
So the function generator still behaves within the specifications.
Many thanks to all who helped me to clarify the phenomenon. :-+
-
To show you the possibilities of the SDG 2042X you see attached a screenshot of an arbitrary waveform I generated following the
instructions from the North American Siglent webpage. (The "Easy"wave X-software is a nightmare!)
It's simple generated with 100 points from an Excelsheet and exported as CSV-File.No further tweaks or so. Maybe there is room for improvements.
Purpose was to overcome the 25MHz-Limitations of the SDG for square waves. The Waveform consists of 10 periods of a Square wave.
As you can see the rise time is now 4ns! But it works "only" up to about 40 MHz. With higher frequencies you get (as expected) a sine-like waveform.
Hope this helps.
scope: SDG2504X+
-
I am now sure to have found the reason for the overshoot probematics.
So it's not the cable or the termination, it's simply the signal amplitude and comes from the SGS2042X.
So if you slowly reduce the amplitude of 2Vpp in steps of 0.1V you suddenly hear a switching noise in the SGS and the overshoot is significantly reduced.
The oscilloscope is able to measure the overshoot (ROV in %).
Siglent gives a value of <3% in the specifications regarding the overshoot.
So the function generator still behaves within the specifications.
Many thanks to all who helped me to clarify the phenomenon. :-+
I know this is a month late but since nobody else mentioned it, that relay click is just the switching in of a 20dB attenuator (and the corresponding boost in the DAC input values) to avoid working with reduced values that would otherwise make quantization noise more intrusive than it need be.
Pretty well all function generators utilise such 20dB attenuators to hold noise (quantization or just good old fashioned thermal noise) at bay at the lower end of the output level range, even the humble Feeltech FY66/68/6900 series switch an attenuator in for levels below the 500mV pk-pk Hi-Z case.
Furthermore, those that can output levels beyond the 5 to 10 V pk-pk (Hi-Z) level, typically maxing out to 20V pk-pk (Hi-Z) level, will similarly switch a linear boost amp in place of that attenuator pad (with a corresponding reduction in DAC input values so as to have another bite of the 'cherry').
Those cheap Feeltech "Toys" include a 12dB boost amp (4 times voltage gain) to cover the range, 5 to 20 V pk-pk Hi-Z case. For frequencies below the 20MHz limit, you can hear the relays click as you transition through their 500mV and 5 volt limits (you may need to disable the beeper in order to hear the relay(s) click).
Although those Feeltech (FeelElec) AWGs had been designed with a 20dB attenuator in mind, they cheaped out on the attenuator resistor network which could then only correctly compensate the shift in DAC values in the Hi-Z case. Bad news for anyone using it to drive 50 ohm loads since it introduced a 2.6dB (or possibly just a 1.6dB - it's been a year since I last looked into this issue) gaping hole in the range of output levels below this 500mV threshold.
Hence rf-loop's sound advice, "Know the limitations of your test equipment!". TBH, I'm surprised he didn't jump in to explain what you had observed. However, it surprises me a little that this change from one 50 ohm impedance source (a high speed opamp with a 50 ohm resistor hanging off its back end) to a 50 ohm 20dB attenuator pad tacked onto the end of that opamp's 50 ohm resistor. Although the attenuator pad has more resistors, it is no more resistive than the single 50 ohm resistance tacked onto the back end of the opamp driving it. I guess the effect is more to do with the relay switch contacts and the difference in signal routing involved.
I've attached the main board circuit diagram of the FY6600 which Derkammi so kindly traced out from his own AWG, to give you an insight into the circuitry around these relays. He'd included the unpopulated U21 and U22 locations which makes the circuit as shown a little misleading.
The designer was obviously an optimist in allowing for the possibility of the chief bean counter relenting on the expense of better components by including the U21 and U22 locations to allow the under-specced dual opamp (U5A and U5B) to be evicted and better single opamps to be fitted instead.
John