A bit like the fashion for middle aged men, my good old Hewlett Packard 8568B got new clothes.
In my opinion, it is one of the masterpieces of Hewlett Packard spectrum analyser design. Even today, the specifications are hard to beat with new equipment in the “amateur price” range. I will return to that.
The 8568A was introduced in 1978, over 45 years ago. The B version was an update of 1985 which mainly consisted of a change of the internal microprocessor. The RF hardware remained essentially the same.
Here are some details and resources about the HP 8568:
In 1988, I started developing software for test and measurement.
More backgrounds are found here: Test and measurement classics
Over time, a weakness for measurement instruments developed and at the same time, it leads to a sort of “accuracy autism” as I would describe it. Modern instruments with loads of signal processing and complex software have gained a substantial market share. Of course, microwave frequencies are nowadays commonly used and most older instruments are simply inadequate for those frequencies. As a result, many traditional instruments are offered for just a fraction for their initial cost.
My 8568B is in a good overall condition and I hope it will stay that way. It is treated with love and is switched on regularly to keep it in good shape.
The analyser display shows the flatness with the HP 8444A Tracking generator. Vertical resolution is 1 dB/division. Start and stop frequencies are 0 and 1500 MHz. The generator range is 500 kHz to 1300 MHz and it is easily seen that the specification of +/- 0.7 dB is matched. The interconnecting cable between the generator (RG58 with BNC connectors) has some loss too and I would have to repeat the measurement with a better cable to reduce the loss.
Modern instruments use “cheat sheets”, for example to normalise the frequency response. The 8568B can be normalised by applying some math to the A and B traces, but the above picture shows the actual response of the hardware.
The challenge for the developers at Hewlett Packard was to create an instrument with excellent specifications and they had to make sure that components like mixer and attenuator were top notch. There are some HP publications with stories about the development process and for me, it illustrates ultimate craftsmanship. These instruments have a soul as opposed to modern stuff which is “sort of OK” and the software takes care of the rest. The same goes for the filters. Easy with DSP, but an incredible challenge with hardware filters.
Do not get me wrong: if I would be working in a lab, modern instruments would be on the bench. They take less space, consume less energy and are very versatile. The software calibration tables improve the accuracy and some top class instruments have specifications which leave the 8568B far behind. But these instruments cost a fortune.
There are two defects, one is a non working left BNC input. This is the DC coupled input, which is fused (thank you HP for protecting it against serious damage) and I assume that the fuse is blown. I will leave it that way, since I do not use it for low frequencies. The AC coupled input starts at 100 kHz.
The second defect is damage to the interconnecting cable. It is electrically still usable, but the mechanical damage to one connector seems to be related to transportation without proper precautions. It looks like the instrument was transported without separating the two units and with the cable attached… I will leave this defect as well, unless it deteriorates in which case I could try to replace the DB connector.
But otherwise, RF performance is well within specifications.
The new outfit
Years ago, I discarded two (ISA slot) National Instrument GPIB interfaces, because computers with ISA slots were discarded as well. But it kept tickling and I decided to order a National Instruments GPIB-HS-USB interface plus some GPIB cables. It took some effort to install the drivers, because I could not find any separate driver software for the USB interface only. Installing a bundle was the only option and it took quite a bit of time to remove unneeded software without removing the drivers.
Some time ago, I stumbled upon the website of John Miles, KE5FX. From what I have seen, John is also one of those who like accuracy and John also wrote some very useful software, which he shares with the community.
In the past, I used a DOS application to convert HP-GL “plotter language” to bitmaps. It should be somewhere in the many backups…
The purpose of it was to save or print “screenshots” displays of instruments, like a spectrum and annotation. Well, the software of John Miles, KE5FX, is a great gift. It saves me a lot of effort to get the DOS stuff going, because it communicates with my instruments and fetches the results as they are displayed. For the time being, I do not intend to do any automated testing anyway…
The KE5FX software even enhances the plots because it fetches raw data separately, which even results in nicer images and annotations. Screen sizes and colour options can be selected and overlays are possible as well. Plots are saved as plot data, which can be loaded into the software to modify the presentation if desired.
The software is here: 7470 HP-GL2 Plotter/Emulator
It is part of a toolkit, which contains loads of interesting software, like a phase noise measurement application and much more.
Phase noise of the HP 8568B
The KE5FX phase noise tool was tried and I was impressed. Here are some results.
Three measurements were taken. The blue trace was with the LPRO-101 Rubidium standard, the Cyan one shows the phase noise with the 10 MHz oscillator in a Marconi Instruments TF2370 spectrum analyser and the green trace was taken with the 20 MHz internal calibration output of the HP 8568B itself.
The LPRO is the best of the three, but is has to be said that the internal reference operates on 20 MHz instead of 10 MHz, which might have some influence. The LPRO also has a higher output level and I will repeat things with 20 dB attenuation between the generator and the analyser.
I do not have a signal source which is good enough to evaluate the performance at 1 GHz, but one thing is for sure: at 10 or 20 MHz, it is still well within specifications.
The same goes for the frequency accuracy. I checked the accuracy at 1.5 GHz after a warm up period of about 4 hours (standby during which the oven oscillator is on). The analyser has a birdie at 1500 MHz, which is related to the 10 MHz internal reference. I made a quick and dirty comb generator to get harmonics of the LPRO-101 Rubidium standard and the error was about 53 Hz, equal to 3,5 x 10-8.
Phase noise is often one of the benchmark criteria and the graph shows that the noise level at 10 kHz offset is about 110 dBc/Hz down. The brochure of the analyser indicates that the close in performance at 1 GHz is roughly 10 dB less and I consider it likely that the 8568B is within spec. Try to find a new instrument that beats this for an “amateur price”.