Review of Agilent 33522B 30MHz 2-Channel Waveform Generator with Arbitrary Waveform Generation

Table of contents

RoadTest: Agilent 33522B 30MHz 2-Channel Waveform Generator with Arbitrary Waveform Generation

Author: Instructorman

Creation date:

Evaluation Type: Independent Products

Did you receive all parts the manufacturer stated would be included in the package?: True

What other parts do you consider comparable to this product?: Sitting beside the Agilent 33522B on my bench is a Tektronix AFG3102 dual channel arbitrary function generator. I will be making occasional comparisons between the two instruments through out my review.

What were the biggest problems encountered?: As is often the case, software installation has been a barrier to full use of this instruments capabilities. I expect I'll have these issues resolved soon, but there is no denying the frustration of a lengthy and unproductive battle with software.

Detailed Review:

Hello Element 14 community members!


There are three parts in my planned review of the Agilent 33522B Waveform Generator.


First is this preliminary peek at the instrument.  I have been exploring the instrument from a first impressions perspective for six days and I have sufficient experience to provide a preliminary review.  Following this preliminary review I will put together a more detailed review, including comparison with the Tektronix AFG3102, an instrument on which I have developed many test scenarios.  Third, if, as a result of my preliminary and detailed reviews, there are gaps that members of the Element 14 community would like to see addressed, I will do what I can to follow up with additional findings, at least in those instances where I have the resources and time to address the gaps.


So, here are my impressions from a preliminary investigation of this instrument:




As can and should be expected from a major supplier of professional test equipment, the Agilent 33522B arrived well packed with Styrofoam cushions.  There were no issues with the quality of the packaging.  Included in the box were three power cords, one for North America and two international versions.  Also inside was a package of two CD ROMs, one containing instrument documentation, the other containing Agilent's BenchLink Waveform Builder Pro (30 day evaluation license on the Pro components).  As can be seen in the photograph below, there was also an envelope containing a certificate of calibration.

Figure 1  Agilent 33522B in box, as received.  Top Styrofoam cushion removed.image

Appearance and esthetics


I like the new instrument design styles that have been emerging from companies like Tektronix and Agilent over the last few years.  Many instruments like ARBs, DMMs, Counters, and Power supplies are being designed in roughly the same form factor.  They often have soft rubber-like bezels, a removable tilting bail, clean, utility driven key layouts, and some sort of display, either LCD, LED, or Fluorescent.  Although not everyone may agree, I like the look of these modern instruments, including the Agilent 33522B.  To me they have a professional appearance and I believe there was some thought put into the layout of the controls to make the end users experience easier.  At the time the photograph below was taken I had connected the 33522B to AC power, but had not powered it up.  You may notice the LED in the bottom left, under the power switch, is glowing yellow to indicate connection to a live AC supply.


Figure 2 Agilent 3522B front panel. Bail removedimage


First power up


I timed the power up sequence for the 33522B.  It took 36 seconds from depressing the power button until a user ready state was displayed on the LCD.  For comparison, the Tektronix AFG3102 on my bench takes 20 seconds to reach the same state. 


The LCDs used on contemporary Agilent test equipment like the 33522B, and on the 53230A Frequency Counter are bright, colourful and crisp.  To my eye they use highly readable fonts (that can be changed to a larger point size if desired).  Unfortunately, my first minor disappointment with this instrument occurred just after the LCD was illuminated.  As can be seen in the photograph below, the LCD is mounted every so slightly crooked, with a small, but annoying tilt to the left.  The image below shows the large point font.  The faint vertical lines that may be evident in the photograph are strictly an artifact of digital photography.  Visually the display is free of artifacts and appears pleasantly smooth and evenly illuminated.



Figure 3  LCD screen.  Notice slight left tilt to the display.image


Ease of use


Normally, I read operation manuals thoroughly before attempting to use a new instrument or tool.  I have found I can save a lot of head scratching and frustration if I first familiarize myself with the operating philosophy the designers had in mind before I attempt to impose my philosophy.  With this instrument I reasoned I could deviate from my established habit and just start using it without reaching first for the operation manual.


This approach worked well.  In the six days I have been exploring this instrument I have been able to achieve every set up I desired without once loading the operation manual on my laptop.  Granted I have only set up simple pulse, sine, square and ramp signals, but the control layout and menu selections were intuitive enough that within minutes I had my intended signals pouring out of the front panel BNC connectors.


Nice features


During my early exploration I tried out one of the features that intrigued me in the Agilent promotional material for this generator.  They claimed you could adjust the bandwidth of the noise signal and add it to other waveshapes.  That was something I had not been able to do on the Tektronix AFG3102, so I thought I'd try it out on the Agilent 33522B.  The Screen captures below show a 1 Vpp 5 MHz sine wave with 1.7Vpp of added noise.  The left image shows time and frequency domain with noise bandwidth limited to 2 MHz.  The right image show time and frequency domain with noise bandwidth limited to 6 MHz.  The horizontal scale for the frequency domain (FFT) is 2.5 MHz/div.


Changing the noise bandwidth was a simple matter of keying in a new bandwidth.  To me, that is a pretty cool feature.





Figure 4  Noise bandwidth adjustment


Well, that is it for the first look overview.  So far I am impressed with this instrument.  However, not everything is going well.  I have been working with Agilent's BenchLink Waveform Builder Pro and have not

been able to get it to communicate over a USB link with the 33522B.  This may be an issue related to the presence of TekVISA on my laptop as well as Agilent's VISA.  Maybe they don't play well together.  If anyone has seen this sort of thing before, please let me know if you got the two VISAs working together.


All the best,



November 4, 2012 UPDATE


Today I have been experimenting with dual channel phase adjustments on the Agilent 33522B.  I wanted to see how easy, or difficult, it is to set up a specific phase delay between the two channels.  For this experiment, I selected pulse waveforms running at 10 MHz with 50.0 ns pulse width at 6.00 Vpp and 8.4 ns rise and fall times (the lowest the Agilent would let me set under these conditions).  The load on both channels was 50 ohm resistive composed to two 100 ohm resistors in parallel.


It was easy to set up the 33522B so that channel 1 was linked to channel 2.  After setting up the parameters described above on channel 1, I pressed the Channel 2 set up button (Figure 5), then selected "More" from the Ch2 output menu (Figure 6), then "Dual Channel" (Figure 7) and finally selected Frequency and Amplitude coupling (Figure 8).



Figure 5  Channel Set up buttons                                          Figure 6  More button on Ch2 Output options menu              Figure 7  Selecting Dual Channel sub menu




Figure 8  Frequency and Amplitude coupled



Once you select "Done", the coupled parameters are prefixed with a blue asterisk on the LCD.  Legend text appears on the bottom left corner of the screen, annoyingly positioned right on top of the word Phase in the display.  This I found a little odd (Figure 9).


Figure 9  Asterisks showing coupled parameters


The unadjusted raw phase delay between the two channels is shown in Figure 10 and is measured to be about 1.4 ns.  Not bad at all.  The phase difference between the two channels can be nulled through a quick twirl of the universal knob.  Once nulled, a new zero point can be set from the phase menu using the "Set 0 Phase" option (Figure 11).


Figure 10 Unadjusted phase delay (1.4 ns)                                                  Figure 11 Setting adjusted phase delay to 0.00 degrees


To examine the precision available in setting phase shift I decided to separate channel 1 from channel 2 by one minor division on my oscilloscope's fastest sweep speed of 2 ns/div.  One minor division at this speed corresponds to 400 ps.  I selected phase from the parameters menu and started rotating the adjustment knob.  at 1.44 degrees I had what appeared to be 1 minor division of separation.


Sometimes when working with a complex piece of equipment, the user uncovers bits and pieces of the designers mind set.  In the case of units for displaying phase shift, the designers of the 33522B believed that phase is represented in degrees and nothing else.  I would have liked the option to display phase in time, but no worries, a quick calculation confirmed that one degree at 10 MHz is about 278 ps.  1.44 degrees is remarkably close to 400 ps, so I am impressed with the available precision in setting phase delays. Well done Agilent.

For comparison,the same sort of phase relationships can be set up on the Tektronix AFG3102.  It is interesting to note that the designers of the AFG3102 believed that phase delay is represented in time and nothing else (Figure 12).  As far as I can determine, both of these instruments have unchangeable units for displaying phase delay - one in degrees, the other in time.


Figure 12  Tektronix displays phase delay in time units


Based on my discoveries today I have upgraded two numeric ratings for the Agilent 33522B.  Specs were sufficient to design with has been elevated to 9 from 7 and Price to Performance ratio was good has moved to 9 from 8.


Still more to come . . .


November 15, 2012 update

A quick update tonight before uploading a full report on the weekend.


I have been exploring the Agilent Benchmark Waveform Builder Pro software that came with the 33522B with two goals.  First, to try out as many of the advanced features as possible before they are locked out with expiry of the full feature evaluation license on November 24th.  Second, to explore Agilent's Trueform technology.


I've learned alot about the Benchmark software and had fun in the process.  I've created complex arbitrary waveforms, DTMF telephone dialling sequences and 16-QAM signals.  I will be sharing my findings this weekend with a complete report including screen captures, photographs and comparisons with the Tektronix AG3102 Arb/function generator.


As a prelude I can honestly report that the Agilent 33522B is a very competent waveform generator with a rich feature set and easy to use PC software, all at an appealing price point.


Check back after the weekend for the details.


All the best,



November 18, 2012 update


After diving into Agilent's Benchlink Waveform Builder Pro application this weekend I have developed a good understanding of its capabilities and would like to share the following findings with the Element 14 community.


To explore Benchlink I decided to use it to generate three waveforms, or sets of waveforms.  First, a DTMF telephone dialling signal, second a 16-QAM pair of signals and third, a complex arbitrary waveform that I could scale in frequency to examine the Trueform technology incorporated in the 33522B.


Generating DTMF dialling signals

This turned out to be a very easy task with Benchlink's equation editor.  I made a table of DTMF code frequencies from reference material located on the web (Figure 13), then opened Benchlink's Equation Editor and wrote separate equation files that summed the row and column frequencies for every symbol in the table.  An example of the equation and waveform for the number "4" is shown in Figure 14.



Figure 13  A table of DTMF symbol frequencies


Figure 14  Benchlink equation editor window showing equation for summing two sine waves to make the DTMF waveform for the number "4"


Using Benchlink's waveform builder tools I was able to concatenate a complete telephone number as shown in Figure 15.  I then copied this file to a memory stick (USB interface is still not working and I think the reason is that I have Tektronix VISA and Agilent VISA on the same laptop) and loaded the file into the 33522B.  The file, when played back as a single manually triggered burst over a speaker, sounds just like a series of DTMF tones should sound.  Although I have not tried this, I think if I held a handset up to the speaker and played the file, a telephone would ring at the other end of the connection.


Figure 15 A complete DTMF 10-digit telephone number assembled from equation files in Benchlink


The only tricky bit to making the DTMF experiment work was to calculate the total amount of time required for the whole sequence in advance.  This value must be entered into Benchlink when opening a new waveform file.


Generating 16-QAM signals


For this experiment, I drew up a constellation diagram of 16 points (Figure 16) then carefully built I and Q signals in Benchlink that stepped through all 16 points in the constellation (Figure 17).  With great excitement I loaded the IQ signal into the 33522B and connected both channels to my oscilloscope.  I set the oscilloscope for XY operation and enabled the outputs on the 33522B.  At this point I realized my oscilloscope cannot display constellation diagrams because it does not have an input to receive an external clock.  Instead of a constellation diagram, I got what looks like an electron cloud around an atomic nucleus (Figure 18).  Regardless, building the 16-QAM in Benchlink was straightforward, even if I don't have the instruments necessary to view the constellation diagram.



Figure 16  A 16-QAM constellation


Figure 17  Time domain view of the assembled I and Q sequence in Benchlink


Figure 18  Constellation diagram fail


Generating complex arbitrary waveforms


I found Agilent's Benchlink Waveform Builder Pro to be very user friendly when working on generation of complex arbitrary waveforms.  For this phase of my investigation I decided to create a complex waveform composed of 10 segments, each 2 us in duration.  The segments are:


  1. The first segment contained two parts, a trigger pulse and a falling ramp.  The trigger gives the oscilloscope something unique upon which to trigger.  This pulse was 4 V in amplitude with 200 ns on time and 300 ns off time.  The falling ramp had a duration of 1.5 us falling from +2V to -2V.
  2. 2us of noise at 4 Vpp.
  3. 2us of ascending stair step, 100 steps 0V to +2V.
  4. 2us of damped sinusoid, damping factor 3.00, +2V to -1V.
  5. 2 us pulse train, 2 Vpp, 10 pulses, 40ns on time.
  6. 2us tri-tone at 6Vpp.  Tone frequencies of 22.5 MHz, 25 MHz and 27.5 MHz.
  7. 2us rising exponential 0V to +2V, time constant -3.00.
  8. 2us falling exponential 2V to 0V, time constant -3.00.
  9. 2us SINC pulse at 2 Vpp with 10 zero crossings.
  10. 2us of SERIAL data at 2Vpp with binary values of 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F.


It took no more than 15 minutes to build the waveform described above while at the same time learning how to use the features in the software. This is a testament to the intuitive user interface and helpful defaults built into Benchlink.  For example, once the user defines a duration for a waveshape and inserts it into the waveform file, the next time a waveshape dialog is invoked, even if for a different waveshape, the previous duration is inserted by default.  The fully assembled 10-segment arbitrary sequence is shown in Figure 19.


Figure 19 A 10-segment complex arbitrary waveform generated in Benchlink


In spite of how user friendly I found Benchlink to be when assembling the complex sequence shown above, there is one shortcoming I encountered that caused some frustration and gave me, yet again, reason to emphasize the importance of keeping excellent documentation.  After building the complex arbitrary waveform and saving it to disk one night, I returned the next day to further explore the 33522B.  As it turns out, I put together the complex 10-segment waveshape "on the fly" without bothering to write details of each segment down.  My assumption was that I'd e able to recover the exact parametric details of each segment within Benchlink.  After searching without success I concluded that Benchlink does not store meta data about each segment anywhere.  I resorted to rebuilding the 10-segments while carefully recording the detail of each segment on paper so I could share the details in this review.


A nice feature to add to Benchlink would be storage of meta data about each and every segment of any waveform that is created, just in case the user needs documentation on what was put into a custom waveshape.  If this feature exists and I've missed finding it, please leave a comment and let me know how to access it.


Editing a complex arbitrary waveshape


To exercise the editing features in Benchlink, I decided to add a noisy Lorentz pulse within the fifth segment (the pulse train).  Using the cursors and the Waveform Math tool, I was able to add in a Lorentz pulse between the 5th and 6th pulses in the pulse train segment.  Then using Waveform Math tools I added noise to the Lorentz pulse.  I was having fun at this point.  I was reminded of the fluidity and rapid feedback experienced when playing with LegoTM .  I could go from concept to visual representation in a matter of seconds.


After playing with waveform math for a while I decided to remove all of my edits except for the noisy Lorentz pulse.  The final waveform was saved and transferred to the 33522B.  I could not cause the 33522B to drop or repeat or misrepresent any portion of the arbitrary waveform over a range of output sample rates from 25 samples/second to 250 M samples/second, a range of 10,000,000:1.  Nice.  However, I thought it would be valuable to compare the 33522B with the slightly more expensive and wider range Tektronix AFG3102 which happened to be on my bench.  Now, I will admit I could not figure out how to use Tektronix's ArbExpress software to generate the same 10-segment complex waveform that was so easily cobbled together with Benchlink, so I simply captured the waveform with my Tektronix oscilloscope, then transferred the captured file to the AFG3102.  You can see that both generators produce a faithful copy of the waveform in Figure 20.



Figure 20  Comparision of Agilent 33522B to Tektronix AFG3102 generating the same arb waveform


I then took both generators through a wide range of reproduction rates to see if either would show evidence of lost, distorted or repeated portions of the waveshape.  The screen captures shown in Figures 21, 22, 23, and 24 show that both generators perform beautifully.  The blue and yellow channels, representing the 33522B (blue) and the AFG3102 (yellow), are placed over top of each other to help reveal differences between the two.  Only at 250 MSa/s in the noise segment detail can we see some fidelity issues emerging.


Figure 21  Noise segment detail at 5 Hz                                                  Figure 22 Serial data segment detail at 5 Hz



Figure 23  Noise segment detail at 50 kHz                                                 Figure 24  Serial data detail at 50 kHz


Just to confirm that DDS generators have limitations with reproduction at certain frequencies, I took the AFG3102 beyond the 250 MSa/s (50 kHz) maximum output rate of the 33522B.  At 60 kHz distortion in the noise segment was obvious.  At 200 kHz, pulses in the serial data segment were badly distorted and by 250 kHz amplitudes in the multi-tone segment were severely attenuated and serial data was unusable.




As a result of the in-depth exploration I conducted with the Agilent 33522B and in comparison with the Tektronix AFG3102 I have emerged with new found respect for both instruments.  I have upgraded my numeric scores for the Agilent in two categories. Product performed to expectations has moved to a solid 10 out of 10 and Demo software was of good quality raises to a 9 out of 10 (just need that meta data feature!).


There are still a few things I'd like to explore, including generating I2C signals with the 33522B and manipulating sampled waveshapes using CSV files in Excel.  The later of these interests me because of a project I was recently involved in that required simulation of fault conditions on a proprietary waveshape.  We manipulated a CSV file of an in spec waveform to simulate min/max violations of 6 parameters then used an ARB to inject the faulty signals into a test jig.  The idea was to validate the software running the jig to make sure it correctly identified all fault conditions.  An Arb is a great tool for this application.  I'd like to see how well the 33522B performs in such an application.


My sincere thanks go out to Element 14 and to Agilent for the wonderful and educational opportunity to participate in this very cool road test.  Road Test is a great program and I hope I've provided an adequate review of this instrument for the community.  I look forward to having more opportunities to explore new technology and sharing my findings with other technicians, technologists, engineers and scientists.


All the best,



November 20, 2012 Post script


There was an interesting side note I forgot to mention in my previous update.  While conducting the tests on the two arbitrary generators with the 10-segment complex waveforms I began to encounter odd behaviors with the wireless mouse on my laptop.  One night the mouse just plain stopped working, so I changed the batteries (they measured low on a battery checker).  The problem seemed to partially disappear, but every once in a while the mouse would lock up or behave erratically.  The next day I noticed, by chance, that there was a correlation between the erratic operation of the wireless mouse and waht I was doing with the laptop.  If I was NOT workig on the Road Test the mouse worked fine.  If I was working on the Road Test, the mouse would sometimes lock up.  I investigated further.  If the outputs on the 33522B and AFG3102 were both active and they were generating the complex waveshape at high frequencies the mouse would stop dead.  If I disabled the output of the generator closer to the laptop (the AFG3102) the mouse would behave erratically.  If I disabled both generator outputs, the mouse worked just fine.  At this point I began to wonder what sort of EM soup was radiating off of my test bench when both generators were driving those 50 Ohm loads.  Whatever was in that soup it was capable of obliterating the bluetooth connection between my mouse and laptop.  I have since decreased the amplitude of my test signals.  Doing so alleviated the assault on the EM spectrum and restored functionality to my wireless mouse.  Moral of the story:  Be aware that generators can produce a fair bit of EMI that can affect nearby wireless technology.




May 19, 2013 Update

I have used the Agilent 33522B generator for a few months now.  TIme spent using the instrument to solve day to day real life problems has further boosted my opinion of this fine instrument.  Today I have another comparative example to share with you.  I will also report on interesting upgrades and fixes available in the latest firmware release (v 2.09).


Identical waveform files - different outputs

I needed to use arbitrary waveform capabilites recently to reproduce custom sensor signals in a controlable way to test the specs of a data acquisition system.  The arbitrary function generators I normally use are from the Tektronix AFG 3000 series.  I compared the Agilent 33522B to a Tektronix AFG3102 earlier in this review.  With an Agilent 33522B now available, I decided to compare the performance of the two generators on the custom sensor signal.


The sensor signals (there are four channels) were sampled at 10 M samples/second in LabView and then converted to individual CSV files. The exact same CSV file of 120000 samples was loaded into Channel 1 of the Agilent 33522B and into Channel 1 of the Tektronix AFG 3102.  The Sync output of the Agilent was connected to the trigger input of the Tektronix in order to synchronize the two generators.  A portion of the output signals from the two generators is shown in Figure 25.


Figure 25


The cursors show that the steps in the Tektronix output are spaced 100 ns apart.  This value corresponds to a 10 MSa/s output rate.  The Agilent waveform, generated at the same rate and from the same file, shows no sign of quantization and is beautifully smooth.  I don't know how Agilent manages this, but I like it!


Firmware update v2.09


The May 5, 2013 firmware release for the 33500 series of waveform generators includes a change to the front panel operation. Phase shift between waveforms can now be selectively displayed in units of degrees, seconds or radians.   Limiting phase shift to units of degrees on the Agilent and units of time on the Tektronix was an observation I made earlier in my review.  Figure 26 shows the menu selections now available for selecting phase units on the Agilent under firmware version 2.09.


Figure 26


Note also in Figure 26 and Figure 27 that the "*Coupled" legend label has been repositioned so as not to obscure other text on the screen.  I had mentioned the odd positioning of this label in my initial review.  If you compare Figure 9 with Figure 27 you will also notice a reordering of the waveshape parameters listed on the left side of the screen.


Figure 27


I have no way of knowing if anyone at Agilent is reading Element 14 Road Test reviews, but it is interesting to me that two very specific observations I made in my initial review have been addressed in the v2.09 firmware update. Perhaps other users of the instrument or Agilent engineers noticed the same shortcommings.  Regardless of how the shortcommings were noticed, I'm glad to see that Agilent is responsive and quick to release upgraded firmware for this instrument.


All the best,


  • I don't know whether to leave this comment here or on your more recent review of the 33622A, but as it more obviously applies to the 33522 I'll put it here.


    The 33522A has a fixed 250MS/s output, though the TruForm approach means resampling and digital filtering is applied before the output. The output samples though should occur at a fixed spacing of 4nsecs. The minimum pulse width is 16nsecs which fits in with this (i.e. 4 sample points). What puzzles me is that the spec sheet gives the resolution as just 100-psecs!


    I'm curious to know whether there is a visible difference in output between pulses of 16 nsecs and those of 16.1 nsecs.

  • You are quite correct about the filter in the Agilent waveform generators.  It is on by default when generating arbitrary waveforms, but can be switched off.  I examine this filter closely in my new Road Test review of the 33622A.  I took a look at the change in high frequency noise caused by turning the filter off on waveforms sampled at 10 MSa/s.  The difference in the spectrum is significant.  The filter I think is a nice option to have, I just wasn't aware that it is active by default.



  • Sorry for the delay in responding to your response, and thank you for coming back so promptly.


    The filter system seems to be an intrinsic part of the TruForm approach, though you can select not to have a filter it makes the jitter worse. More interestingly, on the newer 33622 the sample rate is restricted to 250 MS/s without the filter but can go up to 1GS/s with a filter. I'm not sure why this is the case, perhaps without the filter the overshoot/ringing from rapid steps at high sampling rate cause too much noise.

  • jbridge,


    My recollection is that no output filters were selected on either ARB, so if the Agilent is filtering the input data it might be doing so by default.

    You are correct in noting that the Tektronix ouptut may be more faithful to the original input data.  Good point.  Smoothing should only occur if the user specifically selects it.

  • Good review.


    The smoothing of the steps in Figure 25 seems to imply that the 33522 is applying a filter to the output. Was one of the two arb filters selected deliberately or is one applied by default?


    The output is certainly nice and smooth but perhaps the Tektronix is more closely following the original input arb data. Though the option of applying filters is very useful.

  • Amazing review, after reading I feel like an expert in this. Thanks for your tutorials and images of the product.

  • Very impressive review!

    Thank you!



  • Thanks for the updates Mark, very good data you have collected there.







  • Ah, didn't think of it really image


    But at least your review was very clear to me so it seems to work.


    Glad to hear there is some more results coming, I'm very interested to see them.



    Indeed time is often the problem so updating the review later is always good.