开云体育

Hi Dana,
The Rockland Instruments 7530A and 7530B Spectrum Analyzer plugins for Tek 7000 scopes are outstanding, and unique, Spectrum Analyzers.

In the mid-1970s Rockland Instruments was a small company driven by brilliant engineers who thought outside the box to create state of the art instruments. It was no surprise that Wavetek, which had greater resources and name recognition, bought them several years later to get access to their engineering team.

In 1976 they were among the first, if not the first, to create a spectrum analyzer using an A/D to sample the incoming waveform, convert it into a stream of digital values, perform an FFT (Fast Fourier Transform algorithm) on the data, then use a D/A to convert the results back into the display that appears on the CRT of the 7000 series scope.

HP was not far behind Rockland Instruments. By comparison Tektronix was so deeply committed to swept-IF Spectrum Analyzers that it took over a decade before they realized the advantages of FFT Spectrum Analyzers.

The only limitation the 7530 has is due to the speed and performance of microprocessor chips available in the mid-1970s. The only solution at the time that would be fast enough and precise enough was to create a 12-bit wide CPU using bit-slice microprocessor chips.

The result is a 3-wide 1HZ to 100KHz FFT Spectrum Analyzer plugin for the 7000 series scopes that meets or exceeds the specs of the 7L5 in the same frequency range.

Dennis Tillman W7pF

Hi Dennis,

Do you mind if I copy and paste the above into the Tekwiki's for both?

Seems like good info to have documented while on the topic and available.

I know with the PMI systems... seems all the info in long gone and forgotten other than what I've been able to find with what manuals I've been able to acquire.

Thanks Dennis.

Dennis and Dana,

Wow! Just what I wanted to find. I have a ROCKLAND 7530B that was part of a box of "junk" that someone gave me. II have plugged it into my 7633 and it does create a reaction, but not anything usable. I have the operators manual, but no schematics. I had no idea if this thing was something even worth fixing. From what Dennis is saying, these things are quite good. This thing may be way above my ability to repair, but at least I have a schematic now. Thanks for sharing the manual, schematics and for the history lesson! .

--
Michael Lynch
Dardanelle, AR

Hi Dennis,
I think TekWiki is the best place for it.
Dennis

Hi Michael, Dana, Roy, et al,

Unfortunately the Rockland 7530 has one massive drawback and that is what it takes to repair.

This plugin consists of 6 or 7 (I forget the details) almost full length PC boards that plugin to a motherboard located directly behind the front panel of the plugin. The metal top side and the metal bottom side of the plugin are stamped to create guide slots the boards will fit in as they are inserted or removed from the motherboard at the front of the plugin. Each of those boards has a full row of fingers that plug into their mating connector on that motherboard.

Two of the boards have fingers at their other end as well. One of these boards provides the vertical information and signals to and from the scope so it has to plug in to the interface board of the scope. The other board provides the horizontal information and signals to and from the scope so it also has to plug into the interface board of the scope. It sounds impossible so just In case you aren't sure what I was describing here it is in the simplest way possible: two of the 7 full length PC boards in this plugin have a full set of fingers at each end - one end goes into to the scope's vertical channel (the other one goes into the scopes horizontal channel) and the other end of each board goes into the plugin's motherboard which is behind the front panel of the plugin.

So in order to troubleshoot this thing you need at least two flexible extenders to get power from the scope and to get the plugin sending and receiving signals to and from the scope. That's only the beginning. You still have no way to troubleshoot any of the PC Boards since they are contained in a metal cage and there is no way to get to any of the boards.

You now need to make an extender board that is as long as the plugin so you can troubleshoot one of the PC boards by extending it OUT THE BACK of the plugin. Try to picture this in your mind: You have a 3 wide plugin where each of PC Boards is the full length (less 1 1/2" to allow for the controls of the front panel) of the plugin, and they are the full height of the plugin, and they are in a metal cage you can't stick a probe into.

In order to troubleshoot one of the PCBs you have to put it on the extender you made specifically for this plugin. This extender is just as long as the P CBoards in the plugin. This extender will mate to one of the 7 connectors on the plugin's motherboard which is behind the front panel of the plugin. Once the extender is in place you can troubleshoot one of the PC boards (sort of) by plugging it into the extender. But the entire time you are troubleshooting that board don't forget you have two very wide ribbon cables coming from the flexible extenders in the scope and you will have to reach over those clumsy ribbon cables to get to parts on the PC board you want to troubleshoot.

As if that isn't fun enough you have some very interesting and DIVERSE circuitry to deal with. For example:
* The input section has to have very low noise and have a very large dynamic range.
* The amplitude of that audio signal cannot exceed the 12-bit range of the A to D converter they use.
* 3 different filters (I think they are Hamming, Hanning, and another one) can be applied to the incoming signal to truncate it and make it appear as if it is infinite in length (a requirement of the Fourier Transform) when it obviously isn't.
* From there it is passed to a bunch of bit-sliced microprocessor chips assembled into a 24-bit computer that is running the algorithm that performs the Fourier Transformation on the incoming 12-bit samples.
* The microprocessor needs to be 24-bits wide since a 12-bit wide data word times a 12-bit wide data word results in a 24 bit product. If I am not mistaken a 12-bit word divided by a 12-bit word yields a-yields a 24-bit quotient as well. So far I have not had to troubleshoot the bit-sliced microprocessor in this plugin. Thank goodness.
* One card contains the Dual-Port RAM that is used to store the incoming data, the intermediate calculations performed by the Fourier Transform algorithm, and the horizontal and vertical display data going to the scope. It must be dual-ported RAM since the microprocessor has to juggle real-time data constantly filling one area of RAM, processing already stored data with the Fourier Transform algorithm and saving intermediate results in another area of RAM, writing out the final results to a different area of RAM. The scope needs to be refreshed asynchronously each time there is new data to display but first it has to be converted back to analog.
* The D to A converter has to combine character readout information with the results of the Fourier Transform and convert the entire thing into vertical and horizontal analog signals the scope can display.
* At this point the power supply may seem like it must be the least sophisticated part of this plugin. Far from it. This plugin needs every possible watt it can squeeze out of the 7000 scope. The specifications for the 7000 series were codified in the mid-1960s. The power supply requirements were designed to meet the needs of analog circuit designers. Digital logic was just beginning to appear on the scene so little consideration was given to its power requirements. The only solution was to convert the mainframes +/- 50VDC power and/or +/- 15VDC power to the lower voltage and higher current needed by digital logic and microprocessors. The only way to do this efficiently is with a switching supply. It is a very bad idea to put one of these in proximity to a very sensitive front end like the 7530A or 7530B has. It requires a lot of filtering to reduce the electrical noise from this type of supply. It also requires a huge amount of shielding.

To have any hope at all of ever getting one of these plugins to work you need to have two of them so you can at least swap boards to narrow down where the problem lies to a single board. Even then your chances are slim unless you have experience with the staggering array of technology these Rockland Instruments engineers brought to bear to create this plugin a full decade or more before Tek did. No wonder Wavetek wanted to buy their company.

Dennis

Dennis Tillman wrote:

If I am not mistaken a 12-bit word divided by a 12-bit word yields a-yields a 24-bit quotient as well.

I believe that you are mistaken about this. At most the quotient of a division can only be as wide as the wider of the input operands (given 3 bit operands the two extremes of the computation are 7/1 = 7r0 and 1/7 = 0r7, all the bits wind up in the quotient in one case, and in the remainder in the other). You would need a fill 24 bits if the result of your computation were both the quotient and the remainder, which is often the case in computer architectures (you'll have done all the work to compute both the quotient and the remainder in any case, so you might as well produce them both, and let the program throw away what it doesn't need).

(I'm usually hopelessly out of my depth in these discussions, but I'm a CS guy, and this is right in my wheelhouse)

-- Jeff Dutky

Dannis,

Aside from the physical challenges of working with this plugin, the range of disciplines involved makes it sound absolutely fascinating.

-- Jeff Dutky

Hello,

I followed this discussion and had a look at the 7530A schematic. My job as a physicist deeply involved in electronics is very high resolution spectroscopy of the atmosphere to follow the evolution of its constituents. Since we are using Fourier transform spectrometers with about up to 8 Million points resolution and I am developing these instruments for 3600m altitude observatory, you will understand why I opened this funny schematic.

When I was finishing what is now called my 'Master', my work (=challenge) was to develop a digital filter that was inserted between the then crazy 16bit @ 100Ks/S ADCs to keep only the frequency band of interest and undersample the data to reduce the computation work for the big FFT. At that time, on a 1000F HP mainframe, a one Megapoint FFT took 45 minutes mostly due to heavy hard disk swaps. Memory was small and very very expensive... My digital filter used a specialized computing hardware like TRW 16x16 bit MACs, ECL RAMs, etc... All this under the slower supervision of a Z80 that would prepare the sequencing of the filtering process (convolution) and the necessary operator vector coefficients.

My school friend and now married with my cousin worked on a similar project for making much faster FFTs because 64Kbit RAMs had emerged. He moved to JPL where he is still doing kind of the same job now before the FFT project could find its end. I still have a pile of 64Kbit RAMs and ALUs at hand.

All this to say that I recognize in the Rockland schematic an architecture like the one I devised in 1981. The 'slice processor' looks mostly like an ALU whose control signals come from the sequencer. Sine/Cosine coefficients come from roms and a processor with its attached usual logic prepares all the controls (ALU sequencing, acquisition, display) for the chosen analysis parameters. So, in a sense, the slice bit stuff is not up to the level of a CPU but of a tweaked and specialized ALU. There is therefore probably no assembly language to analyze here but only digital controls for counters, latches, etc. A state machine preprogrammed by the 8080 will provide this bunch of words which is equivalent in a sense to a raw cpu internal microcode but not decoded from higher level instructions.

For the story, when I was making the dgital filter, our experimental Fourier-transform spectrometer used two Rockland 8MHz digital synthesizers to generate
movement control frequencies millihertz apart. This was also an exceptional device for 1980 and is what we now call a DDS that we can buy for one buck or build in seconds in an FPGA with standard IP building blocks.

I remain however doubtful that, at that time with a 12 bit composite ADC and limited memory, the digital approach was giving a real advantage over a 7L5. A few years ago, I had to search for the source of a ghost in our spectra. I tried to use the 7L5 because it is not tied to any of my interferometer controls and works in (slow) real time. I could, just hardly but I could, find this elusive signal when the 'patient' under scrutiny was itself an FFT spectrum analyzer with the usual 2M points @ 18 bits data set. So, a 7L5 remains for me quite a nice and respectable beast at low frequency.

For those interested :

OK, updated for now the following:



Thinking there is more good information to place in this message thread also, along with specs from the manuals to add.

I'll wait a few days for feedback, or edits others might make to the above wiki's, before I add any additional information.

Please reply if you have any comments, suggestions, etc.

Hi Christian,
Thank you for your fascinating comments, particularly since you are so intimately involved in designing similar instruments for your research in Infrared Atmospheric and Solar Physics. As head of the GIRPAS Laboratory at the Université de Liège I can imagine you are quite busy.

Until you mentioned them I had forgotten about the Digital Synthesizers Rockland Instruments engineers designed. These were also state of the art signal sources in their day. I own two of their 1MHz Digital Synthesizers which I used to learn more about spectrum analyzer resolution and phase comparator circuit design. When you have a pair of these synthesizers locked to a Rubidium Frequency Standard it makes it easy to change a 1MHz signal in 1milliHz increments and observe some fascinating details of how phase detector circuits work in slow motion.

The Tektronix 7L5 and the Rockland Instruments 7530A and 7530B Spectrum Analyzers are so fundamentally different that apart from a few basic specifications they can't be compared to each other.

There is something you may have overlooked about the 7530A/B that is not apparent from the schematics which may have been useful when you were searching for the ghost in your spectra that you mentioned. If I remember correctly the specified resolution of the 7530 was 1Hz anywhere in the 100KHz range of the instrument. BUT, unlike Tektronix, Rockland Instruments engineers recognized the importance of designing their instrument to be useful in ways they could not anticipate. They included a front panel connector on the instrument which gave the user access to internal signals. Among other things this allowed the user, by supplying a different clock signal, to increase or decrease the resolution of the instrument. For instance it was possible to increase the resolution from 1Hz to 0.1Hz by reducing the 100KHz bandwidth to 10KHz. I don't think it is possible for a swept IF SA to ever reach that kind of resolution. On the other hand it is quite easy for a Fourier Transform SA to do even better than that.

From your infrared research I was wondering if you are familiar with is the Tektronix J20 / 7J20 Rapid Scan (Optical) Spectrometer. It is capable of scanning the optical spectrum from 250nm (ultraviolet) to 1100nm (near infrared). It can scan any 400nm wide segment of the spectrum in as little as 10mSec. It displays the result as time-resolved spectral power (incident optical power versus wavelength). It has a resolution of 0.25nm.

Dennis Tillman W7pF

Hi Dennis,

This job is indeed very busy and complex. The temperature, ice and wind conditions at 3600m altitude forced me to develop a lot of specific equipment for the task. The problem is not only design, construction and repairs. The need for day-long observations depending on a very unpredictable meteo puts chaos on my agenda and consumes most of the time that would be needed for development. Since 2008, I could implement internet remote measurements that were considered impossible by my predecessor. But covid and the impossibility to reach the lab for repairs put it down since november.

I would like here to bring to the group attention the probable cause of this failure. We had two IP KVMs to control a vintage critical computer. So, one KVM was a spare. The two died within three months and my diagnostic is that the two forgot their firmware programmed at around the same time. One has to be aware that at 3600m altitude the cosmic rays are more intense and speed-up memory loss. This to say that I see a lot of people careful enough in the HP and Tek lists to remove (sometimes desolder) the eproms in their nice equipment to save their contents (if possible and safer, reprogram them to refresh their contents). Now, the situation got worse with the embedded EEPROMs in microcontrollers. In that case, the contents are not accessible and the instrument is doomed when the contents vanishes with time. I already have IP cameras in the lab that seem to have been attacked by this modern electronic Alzheimer. For sure, this is a very convenient unavoidable obsolescence for the sellers. Your motherboards will also forget their Bios in soldered EEPROMs btw. Socketed EPROMS are long gone.

Back to Fourier Transform SA. When you are in the domain, you get quickly the feeling that you can easily buy more resolution by taking more samples, which takes more memory and time. This is the same feeling as an analog SA that needs a slow sweep at small resolution bandwidth. But with FT SA you need a stable source. Otherwise you just would mix-up added noise on top of runing your line profiles. In our case, the target resolution is the one needed to well resolve the thin absorption lines of the atmospheric gases. This means a resolution between 500000 and 1 million. Hence the reason why we usually manipulate 2 million points samples (Nyquist). If the frequency domain of interest is reduced, applying a bandpass digital filter to this set would allow the equivalent of a heterodyne frequency change and the possibility to undersample a lot to speed-up FFT. Otherwise, with a big FFT you can get a 0 to Fmax spectrum at max resolution.

The possibility you mention to slow down the clock of the 7530 is indeed to reduce RBW but at the expense of the maximum input frequency at the instrument input. Because the problem is that you can't increase its maximum resolution which is determined by its memory size. Unless an antialiasing low-pass analog filter was put in front of the S/H, Nyquist would hit you hard and mix-up all your frequencies. ;-)

I never had the chance to see a J20/7J20 and I would have been very happy to play with. It for sure was very compact. Much more than my 4 meter long interferometer.
I remember now that the Rockland synthesizers were operating at 1.6MHz, not 8. I repaired one of them once. A lot of 74xx chips in them and one had failed.

So, group members, save all your Eprom contents when possible and be cautious when buying post-2000 equipment containing soldered or embedded EEPROMs. The time bomb is running...

On Mon, Feb 1, 2021 at 01:11 AM, santa0123456 wrote:

...be cautious when buying post-2000 equipment containing soldered or embedded EEPROMs. The
time bomb is running...

Yes. That is a good recommendation. However, I don't know how to follow it.
If one parses effects into: "short term" exor "long term" OR "acute" exor "chronic"...(to suggest some combination of them)... and then assume the effects santa0123456 observed were short term and chronic...
Then, I don't know how much of the stuff we get (from before the year 2000) came from mountain tops, close proximity to nuclear reactors, prolonged high altitude flights and so forth.
For long term and chronic... say... the U.S. DNA (now renamed) was for many years looking into ways to reduce these kinds of failures... and I'm assuming what they learned was passed on to fabs. I'm also assuming that very expensive equipment from Tektronix meant to support military avionics et. al. ... got a lot of these parts.
Most of us have got bricked stuff, or have bricked stuff... but I don't know how much of "rom rot" (or "recap") is relevant or just YouTube-istic flummox. (I'm not saying it is... I'm not saying it is not: I'm saying, I don't know.)

Gentlemen,

The Rockland 7530A and 7530B are not seriously off topic, because they fit into a Tek 7633 or the 7603, two very popular and commonly found scope mainframe/power supply units

I have one of each Rockland boxes, and a Tek 7603. The 7530A is a working set, but it stressed the 7603’s power supply, such that the system would hang between 20 to 30 minutes after powering on. Eventually, I found that bad or failed solder joints on the several 2N3055 PS main pass transistors was the problem.

The 7530B unit doesn’t work, but after evaluating the innards, those six, (or was it seven) full length boards, seemed almost the same, so I switched one bunch with the other bunch in the 7530A, and the working set still worked! So it’s not the boards it’s something in the 7530B “box” that has failed. I forget which set of boards is currently in the working set, but I never quite finished with the 7603, as it’s so big and heavy, but Iagree, when working, it’s certainly a fantastic setup when evaluating equipment in the audio range.

Steven