Marconi 2023A Signal generator

[Cruisin Marine]

Why on earth would they use a switching PSU as it looks from the piccies? The older one's were analogue and worked well.

[G0HZU_JMR]

It's probably a switcher to save on size and weight and cost. The spectral purity of the sig gen is also extremely good with little to no 50Hz terms on the output. Some of this may be due to the lack of magnetic fields that would otherwise be generated by the type of large transformer that would be used with a linear supply.

One other thing that may be worth mentioning is that the internal AF modulation signals are generated in DSP and these are referenced to the main 10MHz oscillator. So the modulation tones should be very accurate in terms of frequency. Also, the reason there is both a MOD ON/OFF button and a mod SOURCE ON/OFF button is that this sig gen can modulate FM or AM with up to two AF sources. IF you experiment with the MOD button (press it more than once) you will find the option to add a second modulation source tone.

The source ON/OFF button is relevant to the AF tone currently selected. So you can turn off one or both AF tones using this button. It can also do squarewave and triangle wave modulation. The menu system to define all this is annoyingly clunky as it is all done by pressing the buttons in a certain sequence to enable each AF tone. However, all this menu fiddling can be worth it if you wanted to modulate AM or FM with two tones to do IMD testing of an AM envelope detector or an FM detector.

The quality of the AM and FM modulators within the 2023 is also very good. Well above average in this respect and typically much better than the specs given on the datasheet.

[Cruisin Marine]

It's probably a switcher to save on size and weight and cost. The spectral purity of the sig gen is also extremely good with little to no 50Hz terms on the output. Some of this may be due to the lack of magnetic fields that would otherwise be generated by the type of large transformer that would be used with a linear supply.

I notice from Dave's video (which you kindly linked to) the psu is made in Taiwan, so down to cost no doubt, which is pitiful as these units are damn expensive to say the least. Magnetic fields could be lessened greatly by use of toroidal transformers in a linear PSU (and even placed with good thought of design), the size and weight increase would not be much either compared to the overall unit, so all down to cost reduction we can safely assume- sorry, I would expect much better knowing the quite high cost of these units. If I was forking out the amount of money these cost, I would be looking higher in this respect, and I was once a great exponent of Marconi gear.

[G0HZU_JMR]

Whatever the reason for opting for a switcher, the PSU is a bit of a ticking time bomb today although I don't think any of the 2023/4 sig gens at work ever had a PSU failure and there were a couple of dozen of them at one point. All were owned from new. Some were scrapped after being dropped and some got sold off a few years ago. I think there are probably half a dozen of them left in engineering.

When mine failed there was a loud pop and then the sig gen became a silent steam grenade as a large plume of white steam was ejected from the fan at the back and the steam gracefully crept up the wall behind the sig gen as I tried to work out what had happened.

All this steam came from that one popped capacitor.

At work there are also some IFR/Aeroflex 3414 4GHz vector sig gens that also use a switching PSU. There are currently two sat on death row at work, both with failed PSUs. One blows fuses and the other one only starts up when it feels like it. I suspect a startup resistor has aged or failed in that one. I'm not sure if these use a similar supply. I've got an Aeroflex 3416 here but I've not looked inside the PSU area to see the PSU.

[Wendymott]

Hi Peeps...... This forum keeps on giving The U Tube video was VERY en lightning .. only one comment really... the guy giving the report.. Mr Jones does go "on a bit"..
Although saying that.. I have watched his reviews of other bits of kit I bought, namely the Siglent SAA 3000 spectrum analyser. Jeremy, I will order some High temp 100/250 V caps from R/S or other. Ah well... back to "IT"

[Radio Wrangler]

A large number of the wider frequency range 2024 model were dropped on the surplus market by 'The Ministry' several years ago. These can be spotted by their options mix
1) High power output (goes a bit over 200mW)
2) High stability ovened crystal reference oscillator
3) Reverse power protection
4) Fast pulse modulator for radar work.

As Jeremy says, the AM and FM modulation performance is very good. The phase noise is also better than you'd expect.

With all the options going, without a switcher, there would likely be a heat problem in the small cabinet. Also there were threats at various times that things using more than x watts would need to be power factor corrected to meet mains current harmonics rules. x wandered around a bit and drove several firms into switchers when it turned out they didn't need to.

PSUs are the weak points in these instruments. Displays are the next in the pecking order.

David

[G0HZU_JMR]

I think there may be more than one version of the PSU so it's best to check you have the same version of PSU as me. See below for an image of one of the surviving caps from my PSU. This gives the dimensions of the caps used in my PSU.

There's also a high power version of the Marconi 2023/2024 that can output up to +25dBm rather than the regular +13dBm. I've got one of these here but the high power version does have some drawbacks.

The ultimate noise floor performance is generally not as good with the high power version and also the source impedance drops to something like 23 ohms once at or above about +8dBm with the high power version. On both 13dBm and 25dBm versions, when used below 10MHz the sig gen uses a mixer to downconvert to sub 10MHz. This degrades the close to carrier phase noise and also the ultimate noise floor is even worse in this mode when using the high power 25dBm version. Below 10MHz the 25dBm version is one of the worst sig gens I've used in terms of ultimate noise floor. It really is very bad in this respect.

I tried to explain these limitations on the Marconi Groups forum but this was met with some degree of disbelief by a couple of regular experts on there. My guess is they have never used the high power version. I would strongly recommend that the standard +13dBm version is the one to have especially for receiver testing.

There are workarounds with the high power (25dBm) version that can improve the source match and noise floor by using external attenuators or the attenuator hold function. The amplitude flatness vs frequency isn't that great with the 25dBm version although I've never formally tested the 13dBm version for amplitude flatness. Hopefully it will be at least as good.

[Wendymott]

Hi Jeremy and David. All noted thanks . To be honest.. I bought this item on a "whim" thing..Previously I had two 2022's.. Then I bought the Anritsu... and sold the 2022's. Then I realised occasionally it was nice to have two "clean" RF sources. So this 2023 popped up.. I thought.. Im having that. The Anritsu is very noisy mechanically... that fan is like a 747 on full take off boost. So I will favour the 2023 when I get the hang of it.

[G0HZU_JMR]

I think most of the 25dBm versions were 2024 models so yours is probably a +13dBm version. To show you how noisy the 25dBm version is when used below 10MHz I just set mine to the worst (high) power level for S/N ratio. This is +8dBm on the 25dBm version. This is the power level where the last step attenuator is deselected and the attenuator is bypassed. This causes the greatest amount of ALC backoff and so the S/N ratio degrades a lot.

When set to (say) 9.5MHz and a power level of +8dBm the absolute noise floor of my high power 2024 is at about -115dBm/Hz. So this is about -123dBc/Hz. This is really poor and may shock some users.

A good sig gen should manage about -150dBc/Hz here although many will achieve about -145dBc/Hz in a test like this. So the high power 2024 is at least 20dB worse than some sig gens at this test frequency. It is possible to claw back some signal to noise ratio by running the sig gen flat out at +25dBm and then fitting a 17dB attenuator at the output. This should improve the S/N ratio by about 15dB.

If you have the 13dBm version it shouldn't be as noisy as this when used below 10MHz.

[Wendymott]

Hi Jeremy. Mine is the 13dBM version. I have decided, after a look inside, to leave well alone... the caps mentioned are well fixed in as some one said.... If it goes bang in the future.. thats another story. My main question.. Jeremy is this..... on power up..it is going to default setting as in Carrier 1200Mhz etc... My Anritsu on power up goes to the last used settings.. I have looked at the user manual and it says the Power On setting can be changed...is this your experience ?

[G0HZU_JMR]

Within the MENU>>SETUP menu you can select 54 (Power Up Settings) which lets you define the power up setting. Unfortunately you only have two choices.

Option 0 is to let it boot up into the factory default which is what it is doing now.

Option 1 is to ask it to boot to your favourite memory setting number. This could be memory 100 for example.

I don't think it can boot up to the last used settings which is a bit of a shame really. I leave mine in option 0.

[6SN7WGTB]

Just to add to the point made about the change in impedance in the high power mode - I too notice this when the 2025 was connected to a 1meg impedance RACAL 9300 mV meter. As I wound the output up through a certain point the voltage step changed down.

When loaded with a 50 ohm termination it did not (although was halved of course).

However, for the ‘home’ user it’s a rather nice bit of kit. I too bought one on a whim, and buoyed by Dave Jones’ teardown as it is indeed “RF porn”…! Mine is indeed an ex-MoD unit, with carry case, full printed manual, mains cable (unused with tie still on it) and bag of two spare fuses!

[Wendymott]

Hi Jeremy and 6SN7GT..... Ok Jeremy.. a bit unfortunate...... I will set it to Option 1 and see how I get on with it. I Like the "RF Porn" term

[G0HZU_JMR]

To show how good the quality of the NBFM and AM modulation is for the 2023/4/5 I set up the 2024 to two tone modulation for NBFM with 2.5kHz deviation per tone. See the audio spectrum when demodulated the output using my Racal 9008M modulation meter. I fed the demodulated AF out from the back of the 9008M into a spectrum analyser.

I did the same for AM modulation and in this case each mod tone had 30% modulation. The IMD3 distortion of the demodulated FM is not visible so this is really impressive. There is a small amount of distortion for the demodulated AM but I'm not sure if this is a limitation of the 2024 or the Racal 9008M. It's still an impressive result.

[Wendymott]

I am beginning to like this generator. Just getting used to the menu system.. One thing I need to know.... If I use the 2023 as a low signal source.. say 1 uV @ 145 Mhz.. and the Anritsu MG6342 as a high level source to check near frequency performance,.. how should I couple the two... I guess with a VHF Torroid, and a secondary winding to the receiver, but I dont want to "feed back" a large signal into the 2023. Jeremy, your valuable wisdom is needed here

[G0HZU_JMR]

You could experiment with a Sontheimer directional coupler using a large twin hole ferrite bead but I'd be tempted to stick with the classic 6dB combiner as found in the ARRL handbook.

The image below is from the 1989 handbook. I still have and use the one I made well over 30 years ago. It works from about 1MHz through to a couple of hundred MHz with really good isolation. I can't remember for sure but I probably used a classic FT37-43 NiZn ferrite toroid. I made another one a few years later and I used an MnZn ferrite and this version worked down to just a few kHz. I also still have this version somewhere.

Both were made in a hurry and are very ugly in terms of construction but they have stood the test of time.

One thing to be wary of is the noise floor of the sig gen that provides the large signal. Often the noise floor of the sig gen will be the limiting factor if you wanted to do a reciprocal mixing test of a high performance receiver at about 10-100kHz offset. Often, it's best to use an LC oscillator (or a high drive crystal oscillator) for the big signal as this should produce a very low noise floor at 10-100kHz offset.

Your 2023 will typically manage a noise floor of -150dBc/Hz. This is generally about as good as it gets for a lab sig gen. An LC oscillator can easily achieve -170dBc/Hz and a crystal oscillator can get close to this if the crystal is driven fairly hard. The crystal oscillator has the advantage in that it can achieve low phase noise even at 1-10kHz offset.

[Cruisin Marine]

That ARRL Hybrid combiner may be in several issues of the ARRL Handbook, it was in the 1981 release with a red cover (page 16.37).
I built one back in the 80's and have used it ever since - it gives about 40-50 dB isolation with 6dB loss, it is extremely good.

[djsbriscoe]

Are these 2023/24 generators easy to fix? I might pick up a fixable one at some stage over the next few years. I remember using one in my last job.

[Wendymott]

Hi Jeremy. Thanks for the info.... Im not into "serious" measurements.... I just was curious if I decide I need to test my 2 M 6 M front end.
DJS. Looking in mine... I would say no.... not like the 2022, whom were well known for the ceramic IC's to fail. Apparently the 100 uf/250 V caps in the PSU are a bit of a problem... but I am crossing my fingers on that one.

[G0HZU_JMR]

Here's a few old phase noise plots of one of the works Marconi 2024 sig gens. This is a +13dBm version and the plots are taken at 10MHz, 50MHz and 144MHz in case this is of interest to you Wendy for your 6m and 2m projects.

When the sig gen is tuned below a carrier frequency of 10MHz, the phase noise will probably be slightly worse than the phase noise plot at 144MHz because the sig gen mixes down from VHF for all signals below 10MHz. So the phase noise here is only as good as the phase noise up at VHF. It's still better than a lot of other sig gens although the noise floor probably won't be as low as it is on the other plots. You can see that the other plots get close to a -150dBc/Hz noise floor at wider offsets.

The 10MHz plot below shows the lowest frequency that the sig gen achieves really good phase noise. Tune it down just 1kHz from 10MHz and the phase noise will be similar to the 144MHz plot but the ultimate noise floor could be maybe 10dB higher at about -140dBc/Hz for carrier frequencies below 10MHz (at a guess).