Marconi 2018

[synx508]

I bought two Marconi 2018s at the Newbury rally, both not working and a few keys missing from the keyboard - though not the same keys.
One was £45 sold as "stopped working during storage" and the other was £20 and sold as incomplete.
The incomplete 2018 was missing its Frequency Standard board, but the oscillator was still present. The standard board divides the oscillator by 10,000 to provide a 1kHz reference to the synthesizer and handles the switchover to external clock source, too.
My first job was to see how badly messed up this incomplete unit was, by substituting the Frequency Standard board from the supposedly working one. This didn't make it spring into life. The computer did not appear to be communicating with the other boards, but was running and starting up.
Instead of swapping the board back I decided to board swap it until it worked but first I'd check the capacitors in the PSU.
Sure enough all the small Philips blue capacitors (22µF and 4.7µF) were well down in value (3µF and 1µF respectively), so I changed them. I also took out the chunky 15mF 16V smoothing capacitor for the +5V rail and somehow managed to misread it as 18.5µF, so I changed it. Later I put my glasses on and discovered that it was actually 18.5mF, I'd forgotten that the small tester I was using switches to milliFarads. I had tested the small capacitor on my Wayne Kerr B605 so I'm not sure why I didn't test the big one on that, too, but it was hot and I was tired and my vision was blurry.
I decided that as the computer comms problem was preventing anything from happening and limiting my diagnostic options I'd switch back to the other supposedly nearly working unit. This turned out to be not really the case, in many ways, but the computer *was* working and so was the front panel and something was appearing on the display. The output, though, was nothing because the reverse power protection was triggering once the computer finished booting. Removing the connector that contained the input to the RPP measurement diodes allowed it to boot but it was clear that the display wasn't right.
I decided to use the rebuilt-PSU generator as the project and the "working" 2018 as the donor, the deciding factor being the "working" PSU which had been modified for no obvious reason so the +5V adjustment range started at 5.25V and went up to around 7V, which I thought may have damaged some TTL or one of the divider ICs. This turned out not to be a particular problem but a duff keyswitch sealed the fate of the front panel and later I discovered that the frequency display had some bad segments (could be the decoder/driver chip).
With the Frequency Standard board now moved back to the more broken of the two I found that it would occasionally boot. Changing the CPU and reseating the EPROMs seemed to make it better but it always started up with an error indicating a bad checksum in the EAROM.
I also had no synthesizer lock, but it was switching between oscillators. There was no output on the front panel, though, only at the intermediate point where the RF comes from the synthesizer into the AM/filter/amplifier box. I took the easy way out at this point and substituted two divider boards in the synthesizer which brought back the fine steps but only above 32.5MHz. This meant that the BFO circuit was not doing its thing, so I changed that and got 80kHz to 32.5MHz back. The attenuator on this one seemed to be fine, though the output levels were all wrong, it was at least going through 10dB steps as it should.
Then I tried FM, but there was no FM at all. It seems that with an empty or corrupt FM calibration table the generator can't do FM. Additionally there seemed to be a problem with the FM/modulation oscillator board, so I swapped that over, too. After a very long calibration process with my trusty (slightly broken) 2305 I had working FM.
Now it was time to look at the output level, which seemed not awful on the BFO output but very poor at the UHF end, even with the calibration cranked up all the way. The output amplifier board had been modified due to excessive gain. Rather than back the mod out I just changed it and this brought back full level adjustment and amplitude calibration was a success with plenty of gain in hand. I didn't realise before, but these generators can easily do +20dBm with a bit of deliberate mis-calibration. AM calibration also worked but I wasn't able to stop it being a few percent over at UHF, so I may revisit that soon.
The front panel needed attention so I put together the best switch caps from both to give it a full set. Two LEDs were also not working, the AM indicator and the 6kHz modulation oscillator indicator. I changed both multiple times, turns out the + symbol on the silkscreen is on the cathode. No idea why but not the first time I've seen this on test equipment. I didn't have any 2mm LEDs for the 6kHz indicator or proper yellow 3mm for the AM switch, but I managed to pull the 2mm LED from the other front panel and have used a sort of orangey LED instead of yellow for the AM switch. It's actually the same colour as the switch, which almost made me change all of them…

I'm now trying to get the donor unit working to some extent, as I am sure I can bodge a Frequency Standard board quite easily and I'd like to get board level fixes done where possible.

I had originally intended to sell the 2018 at McMichael but it wasn't ready in time - just the LEDs, really, but I'd worn myself out and had little sleep, hadn't managed to get to to the storage unit to grab some other items I wanted to sell. I may make it to MK to sell, though. I'm definitely keen on getting rid of my 853/8558 (with bonus broken 8559 and incomplete 8557 in flatpack form), possibly my 3325B, because it's a hateful thing that breaks whenever I need it, some Fluke 8502 ac meters, maybe my RACAL 1992 and 9300B and a few boxes of components that I'll never use. Not sure it'll fit in the Fiat 500, though…

[Zenith]

Interesting and a post worth noting for the future. I eyed up those two 2018s at Newbury and heard the patter about one being incomplete, but the dearer one only needed the boards wiggled. I was tempted. There was another 2018 on the stall next door for £70, as I recall. I don't know what claims the seller was making. There was a fourth in the back of an estate car.

I have a working 2019 which I use, and a 2019 I bought as a spare, which is mainly working but has problems. The keys and LCDs on those act up and the frequency references very likely weren't great when they were new, and have not aged well. They are nice sig gens, but they are big, which is part of why I could resist the pair at Newbury. Had they been 2019s it would have tipped the balance.

[synx508]

It's quite tempting to populate the doubler and flip the bit in the configuration that tells it that it's a 2019 and not a 2018. The RF amplifier is capable of the full frequency range and there's an extra calibration point for setting the levelling from 520 to 1040MHz.

[Zenith]

It's amazing the number of things where the much more expensive de luxe high performance version, is very little different from the standard one. e.g. Siglent and Rigol TE, where the right licence codes turn the bottom of the range into the top of the range, with all options available.

I suppose the electronics industry has always sold a guaranteed specification, and how they do it is their business. The earliest example I can think of is from around 1930 when a U12 and U14 rectifier looked the same but the U12 had a specified voltage of 350 and the U14 500. Now there may have been differences in materials, or U12 was only tested to 350V, or all U12 and U14 tested to well over 500V, and they marked them differently to suit market demands.

[EC8010]

Tektronix and HPAK have been doing it for decades. I bought my Tektronix TDS3032 oscilloscope in about 2000 and on one of the firmware updates I discovered it was sampling at 5GS/s rather than maximum of 2.5GS/s that I'd paid for. The next firmware update "corrected" that. If you buy a bottom-of-the-range Tektronix, HPAK, or R&S oscilloscope, you can upgrade its bandwidth from 300MHz to 500MHz or 1GHz with a licence code; the top-of-the-range front-end is there already. Likewise, you can upgrade the maximum record length with a licence code; the memory is there already. They just make one oscilloscope, then hobble it for different price points. I can see that it makes for economies of scale to make a single instrument and tailor it by firmware, but it raises serious questions about the actual cost of production, the selling price, and what you're paying for. I can't help feeling that the major manufacturers set prices by what the market has historically withstood and that they operate as a cartel. Have you noticed that you have to pay a premium to get AC RMS as an automated measurement or the ability to average across FFTs? Neither is an expensive calculation, but oscilloscope ranges have been deliberately structured in that way. As the engineers who accepted that (analogue) bandwidth genuinely cost money to make die away/retire, that tolerance of artificial pricing is withering and much cheaper oscilloscopes are probably bought by Generation Z. That's all well and good if those cheaper oscilloscopes are genuinely as good. The problem there is that not all oscilloscope deficiencies show up in the specifications, so a cheap Siglent (for instance) produced a very poor FFT compared to my TDS3032 because of sampling clock jitter. Real-time capture rates can be slow because the computer is too slow to keep up with data production from the front end. By setting their prices artificially high, the major manufacturers may well be able to please the short-term aims of their bean counters but it's a poor long-term strategy and encourages design/production inefficiencies that will lead to their eventual collapse. And that's a bad thing.

Turning to U12 and U14 rectifiers, have a read of "The Saga of Marconi Osram Valve". It's a very good book written by two engineers who worked there and there are lots of snippets about how valves were made. In one of the transmitting valve sections, there's a piece about how valves were cleaned up after manufacture to allow them to tolerate higher voltages. The Marconi valve book is best read in conjunction with "The Oxide-coated Cathode" (Vol. 1) by Wagener which is quite hard to find as a paper copy, but both volumes are on the World History site that has Wireless World.

[tggzzz]

Tektronix and HPAK have been doing it for decades. I bought my Tektronix TDS3032 oscilloscope in about 2000 and on one of the firmware updates I discovered it was sampling at 5GS/s rather than maximum of 2.5GS/s that I'd paid for. The next firmware update "corrected" that.
...
but it raises serious questions about the actual cost of production, the selling price, and what you're paying for.
...
By setting their prices artificially high, the major manufacturers may well be able to please the short-term aims of their bean counters but it's a poor long-term strategy and encourages design/production inefficiencies that will lead to their eventual collapse. And that's a bad thing.

They aren't alone, of course.

In the 60s IBM sold their 360 mainframes using a very wide range of implementation technologies, and promising an upgrade path as customer's requirements increased. IIRC the rule of thumb was that cost=k*performance².

In some cases the performance was increased simply by removing a link on the backplane.

[Zenith]

Tektronix and HPAK have been doing it for decades. I bought my Tektronix TDS3032 oscilloscope in about 2000 and on one of the firmware updates I discovered it was sampling at 5GS/s rather than maximum of 2.5GS/s that I'd paid for. The next firmware update "corrected" that. If you buy a bottom-of-the-range Tektronix, HPAK, or R&S oscilloscope, you can upgrade its bandwidth from 300MHz to 500MHz or 1GHz with a licence code; the top-of-the-range front-end is there already. Likewise, you can upgrade the maximum record length with a licence code; the memory is there already. They just make one oscilloscope, then hobble it for different price points. I can see that it makes for economies of scale to make a single instrument and tailor it by firmware, but it raises serious questions about the actual cost of production, the selling price, and what you're paying for. I can't help feeling that the major manufacturers set prices by what the market has historically withstood and that they operate as a cartel. Have you noticed that you have to pay a premium to get AC RMS as an automated measurement or the ability to average across FFTs? Neither is an expensive calculation, but oscilloscope ranges have been deliberately structured in that way. As the engineers who accepted that (analogue) bandwidth genuinely cost money to make die away/retire, that tolerance of artificial pricing is withering and much cheaper oscilloscopes are probably bought by Generation Z. That's all well and good if those cheaper oscilloscopes are genuinely as good. The problem there is that not all oscilloscope deficiencies show up in the specifications, so a cheap Siglent (for instance) produced a very poor FFT compared to my TDS3032 because of sampling clock jitter. Real-time capture rates can be slow because the computer is too slow to keep up with data production from the front end. By setting their prices artificially high, the major manufacturers may well be able to please the short-term aims of their bean counters but it's a poor long-term strategy and encourages design/production inefficiencies that will lead to their eventual collapse. And that's a bad thing.

It's a marketing strategy that may appear dishonest but has stood the test of time. The manufacturer sells you a specification and is willing to stand by it. They mainly sell to other manufacturers, or the military, or other large organisations. Does anyone seriously want the prospect of facing a disciplinary hearing, or court of law, or court martial because they used a part which was under spec (but their testing showed, and it was common knowledge they were the same) failed and could be accused of causing a disaster, to save a few quid, not of their own money? Large companies dealing with other large companies are in a position to negotiate these things.

I've no idea how much Siglent and Rigol sell to large organisations vs amateurs. I assume that professionals buy a piece of gear, use it and it goes off for calibration as a matter of routine. Then after a few years, it's sold off. They are not going to mess about with unauthorised modifications because they are not. Amateurs, OTOH are not going to buy this stuff unless they know they can adjust the lowest price option to be the highest priced option, and they'll take their chance on whether it remains in calibration and the effect on the warranty. In some cases there's an official upgrade path from the lowest to the highest, and in others there isn't, so there may be actual differences in the hardware, possibly revealed by factory testing.

I guess that HPAK, R&S, Tek etc, mainly sell to to large organisations, which are not cost sensitive and where they have lock in anyway. I assume their products are better, and for the price difference, they certainly ought to be.

I was always intrigued by 1N4001 to 1N4007 rectifiers which came out in the early 60s. I wondered if there was ever a difference, or whether they'd all pass the tests for a 1N4007. For years there's been little or no difference in price, and you may as well use 1N4007 and forget the rest.

[EC8010]

I assume that professionals buy a piece of gear, use it and it goes off for calibration as a matter of routine. Then after a few years, it's sold off. They are not going to mess about with unauthorised modifications...

I don't agree with your argument about pricing, but I accept that it's a valid argument. When you paid extra for more bandwidth on a valve oscilloscope, it was because it genuinely cost more to get that bandwidth, and that was honest trading. That's not true now it's all on a chip. I have learned when buying stuff for myself to buy the bottom model of the highest range I can afford, so my Tek oscilloscope is a paltry 300MHz when it could be 1GHz. I rarely use it at time bases faster than 1s/div. I suspect Rigol are more targetted towards amateurs.

Depends. At the Beeb, we bought stuff and it was never recalibrated in its entire life. But to be fair, the standards for measuring video were quite low, so superlative accuracy wasn't necessary. When I was in teaching, the main criterion was robustness. In my final job, we did send some stuff away to be calibrated, but it was only one out of each type of instrument. Our lab probably had about twenty 6 1/2 digit DMMs, three electrometers, a current source, one LCR bridge, and a few odds and ends, so one of each went away for annual calibration. I've never worked in a lab where equipment was sold off; if it went, it went in a skip and it would be at least twenty years old by that time. Probably with scorch marks somewhere. Agreed that it was uncommon to modify commercial equipment. In final job, if a standard piece of equipment couldn't do what we wanted, we designed and made something from scratch.

Point noted about 1N4001 vs 1N4007. I've certainly got some 1N4007, but whether I have anything lower voltage, I've no idea. If I have, I may do some measuring to see if there's genuinely a difference.