Generators
A look at level stability
Fluke 510A 2400 Hz
Fluke 510A 400 Hz HP 3320B NI USB-4431 Philips PM5191 Simple Wien bridge SRS DS335 R&S SPN KH 4200A HP 3326A Wavetek 29 HP 4204A Philips PM5141 B&O TG8 |
3.15 ppm
3.18 ppm 5.89 ppm 17.0 ppm 23.5 ppm 25.4 ppm 30.9 ppm 123 ppm 211 ppm 228 ppm 236 ppm 250 ppm 754 ppm 997 ppm |
The figures on the left show the relative standard deviation of the output level over 24 hours for selected low-frequency generators measured with the Fluke 8506A Thermal RMS Digital Multimeter. The frequency is 1 kHz unless noted otherwise.
Obviously, the Fluke 510a AC references come out with the best standard deviation in this test. The old Hewlett Packard 3320B turned out to have a quite low standard deviation, thanks to its leveling feature. The "simple Wien bridge" is a DIY oscillator based on a full-wave active rectifier, an AD580KH voltage reference and an integrating control loop, using only non-precision components. The two generators with the worst standard deviation were both suffering from unstable push-button switches in the output circuitry, which caused occasional level reductions, sometimes slowly, and sometimes abruptly. Some "push-button massage" was required to maintain a smooth level. Indeed, old push button switches require some attention, and an occasional cleaning, if at all possible. |
A look at synthesized generators and systematic frequency errors
Not all synthesized generators are created equal, and a high number of digits on a front panel may be deceiptful, even when the generator is connected to an external reference. Below I take a look at some DDS and Fractional-N designs. In some cases, an old generator could be more suited for your task than a new.
The 10 MHz ChA output of the 3326A compared to its 10 MHz reference output over 2 hours. Click on the picture for a larger version.
|
Hewlett Packard 3326A Two Channel Synthesizer
There are indeed other two-channel function generators than the old 3326A on the market, and they are even not as bulky, heavy and noisy. On the other hand, the sturdy build and the modular design of the 3326A is appealing, so if you have space on your desk, and if you think you can live with the noisy fan, it's worth considering provided the price is not too high. The advantage over the single-channel 3325A, by the way, is that the 3326A allows you to do fault-finding through module swapping. There is a feature of the 3326A which you may find particularly interesting: When connected to a 10 MHz reference, the 3326A will, thanks to its Fractional-N design, give you the exact frequency you have entered, which not necessarily will be the case with many new DDS generators. The reason is that the phase incrementation in a DDS generator has to be truncated if the sample frequency of the D/A-converter has a "nice round number", such as an integer times 10 MHz. To get a correct frequency from a DDS generator, with no unintended phase wander, the internal reference must have a frequency which is related to 2 to the power of N. Further information on the topic is found in the description of the TG5011A below. The 2 phase mode of the 3326A can be useful, but do not expect that the phase between the two channels is maintained if you change the frequency. In fact, you need to run the internal calibration routine if you change the frequency, or if you toggle between sine or square wave, in order to avoid a phase offset error. Enable the auto cal mode so that the 3326A makes the required adjustments when you make changes. |
Picture to be added
|
Hewlett Packard 3325B Synthesizer / Function Generator
The 3325B is an update of the widespread 3325A, easily recognized by the rubber buttons that replaced the well-known tactile plastic buttons. Besides this, HP carried out a number of design/reliability changes. If you need details on the differences between the two models these are described in an internal note included in the 3325B service manual which can be downloaded from Keysight's web pages. Check out pages 198 through 201 in the pdf file. One of the advantages of the 3325B is that you get the frequency you type in (just like the 3326A above) without any systematic errors typically found in modern DDS generators. An advantage over the 3326A is that the 3325B goes up to 21 MHz, while the 3326A only goes up to 13 MHz. Actually, the 3325B supports frequencies up to 60 MHz, but the signal is then located on the AUX output on the back panel. The IEC power inlet requires a word of caution: It's one of the infamous Schaffner models that are reported to suffer from spontaneous decomposition of the noise suppression capacitors, followed by the emission of smoke with a truly foul stench. And sure enough, this happened one day to my 3325B...!! The badly smelling goo emitted from the main filter had coated various connectors and cables, including large parts of the GPIB board, and the RS-232 connector. I dismantled the back panel and cleaned up the mess. This had to be done outside, due to the stench, while wearing nitrile examination gloves. After the cleaning session (which lasted a couple of days!) I replaced the IEC power inlet, the fan, the RS-232 connector including its Molex KK254 connector, the Grayhill DIP switch on the GPIB board, and every Mate-N-Lok connector housing. The moral of all this: If you get a 3325B, just replace the IEC inlet with a new Schaffner model FN370. Don't wait for the IEC inlet to self-destruct. Mind you that I do not have any second thoughts about using Schaffner again; It's a well-reputed company which had the misfortune to use capacitors that turned into exploding smoke generators after a couple of decades in operation. |
Phase wander at different frequency settings of the TG5011A (click on the pictures for larger versions). Note the different y-axis on the three graphs.
|
TTi TG5011A 50 MHz Function/Arb/Pulse Generator
This is a fine example of what a modern DDS generator can offer, and at an attractive price. The width of the phase accumulator (and thereby the frequency resolution), the D/A-converter update rate, in combination with the rich feature set, make this generator a very attractive candidate for an all-around generator for the laboratory. Speaking of prices: Watch out for the highly elevated prices of the TTi generators on your favorite auction site. You will find that the price offered by the authorized dealers often is MUCH lower!! The 5011A has a 10 MHz reference input so that the frequency can be locked to an external frequency reference. However, the internal 125 MHz D/A update rate will for most of the frequencies entered call for a truncation of the phase accumulator incrementation. As a result, you will rarely get the frequency you think. There will be an offset, equal to a phase wander, which depends on the chosen frequency. For a lot of applications, this does not matter, not even the slightest. For a few specific applications, it definitely will. For instance, imagine you would like the TG5011A to output a 5 MHz signal when connected to a 10 MHz reference. The frequency you get will be about 2e-14 off, which does not seem like much, but the systematic phase wander would be an issue in a lot of setups. At lower frequencies the effect of the truncation becomes more evident. How about a 1 pps signal? Well, the output will be 0.36 ppm off. In view of this, it's fair to say that the systematic error caused by the phase word truncation effectively prevents generators such as the TG5011A, in spite of its otherwise fine qualities, to be used for applications that require a fixed timing relation between the reference and the output. The upper two graphs on the left show the phase wander of the TG5011A compared to the output of the HP 3326A, when using the 10 MHz reference output of the TG5011A as the reference input for the 3326A. Examples are given for 12345 Hz, 60,000.001 Hz and 198 kHz (upper graph), and for 1 Hz (middle graph). The data were captured over 2 hours for each setting. The lower graph shows the output of the TG5011A when set to 10 MHz compared to its own 10 MHz reference output over 48 hours. The phase wander measured in all cases is consistent with a 48-bit accumulator in combination with a simple truncation of the required phase accumulator incrementation down to the nearest integer. |
Wavetek model 29 (TTi TG1010) DDS Function Generator
The model 29 from Wavetek shown here is a relabeled TG1010 from TTi. Both are now discontinued, but appear regularly on the used market. With a 38-bit phase accumulator the resolution of the generator is not as fine as the later TTi models, and the frequency is limited to 10 MHz. The reference input is not made for 10 MHz, but for the internal sample rate of 2^38 * 0.1 mHz (27.4877906944 MHz), which at first seems like a drawback. However, this can be turned into an advantage, and is the main reason why I highlight this generator: By using a sample frequency of the type M * 2^N the generator does not need to apply any truncation of the phase accumulator word, and lets you get the exact frequency you have entered when the generator is locked to an external reference. So what to do with a laboratory reference of 10 MHz? The answer is an external reference clock converter, which converts a 10 MHz laboratory reference to 2^38 * 0.1 mHz required. Thanks to such a converter, you may set any frequency on the generator without risking an offset or phase wander. |
Time difference between 3326A ChA and the PM5191, both set to 1 Hz, when using 3326A ChB set to 8.589934592 MHz as the reference for the PM5191.
|
Philips PM5191 Programmable Synthesizer / Function Generator
This is another example of a generator in which the internal update frequency of the D/A-converter is not a "nice round number", and the reference input frequency is not 10 MHz. The PM5191 uses a 40-bit phase accumulator and an update frequency of 2^33 mHz = 8.589934592 MHz, which is also its reference input frequency. With an external converter from 10 MHz to the required 8.589934592 MHz reference frequency, you may set a frequency on the PM5191 and get exactly that on the output. As the converter, I use my own design as shown on the timing pages. As an alternative, you could use a generator such as the HP 3326A. If you feel tempted to look for the PM5191S generator variant with 10 MHz reference input, just consider that the circuitry included in this version converts the 10 MHz input to a 10 MHz * 9217/10730 reference frequency, which differs from the nominal 2^33 mHz by 0.02 ppm. The PM5191 has a mere 9-bit D/A-converter, so do not expect stellar performance. There will be jitter, and in particular at low frequencies. The PM5191 and generators of similar design should not be put in use for producing low-jitter sample clocks or perhaps a 1 pps signal. To lower the jitter, an external low-pass filter and shaper would be required, but then a dedicated, fixed divider would be a better idea alltogether. |
Adret 3310A Frequency Synthesizer
This is an example of a high-resolution synthesizer before DDS became mainstream, and which did not resort to a fractional-N design, for instance, as found in many HP products. The 3310A is designed as an elaborate multi-decade indirect synthesizer where there's a PLL for each power of 10 in the frequency setting. This generator is really something if you're into analog design with lots of mixers, bandpass filters, and that sort of things. One advantage of the design is that there will be a correct integer ratio between a 10 MHz reference and the frequency of the output signal. In other words, there are no systematic frequency offsets due to truncation errors like in most DDS designs. From the schematics of the 3310A it can be seen that the PLL which locks the 3310A to a 5 or 10 MHz reference does not include an integrator, so the phase of the master oscillator may wander a bit with respect to the reference, thus degrading the stability. A better design would be to have an integrator, which then is nulled if there's no reference input. The Adret 3310A reminds of the Hewlett Packard 3320B, as both are indirect synthesizers with ALC, but the synthesizer designs are different. The 3310A excels in a 300 Hz to 60 MHz frequency range with 0.1 Hz resolution at all frequencies. The non-harmonic spurious contents of the Adret 3310A are specified to be 75 dB below the carrier, or better, wheras the HP 3320B has a specified spurious level of 60 dB below the carrier. The 3310A was not working when I got it, but I soon discovered that 4 tantalum capacitors were shorted in the power amplifier module, and that one 10 nF ceramic capacitor in the oscillator in one of decade modules was intermittently turning into a resistor of about 500 Ohms. After these components were replaced, the 3310A worked again. I also replaced the Schaffner IEC mains power inlet to avoid a sudden (and smelly) decomposition of its old noise suppression capacitors. For safety reasons, I relocated one of the capacitors in the power supply, as it was almost in physical contact with the terminals of the fuses. On the picture you can see three electrolytic capacitors with red sleeving, next to the power supply module on the upper left. There's a capacitor which is stacked on top of the other two capacitors, but this capacitor was originally positioned (too) close to the fuse holders. |