TX Noise

[M0DTS]

Hi All..

has anyone else experienced (like me!) some wideband noise either side of the TX frequency like the image below?
This was with both I and Q drive pots at minimum.

I cured it by replacing C22/C25 with 100nF, maybe be pickup from IQ drive on the board onto the bias supply.

Another note just to add is that you can use this: LMH6658MA opamp from Farnell ias a direct replacement for the AD828 at a significantly lower cost!

Rob
M0DTS

[w0fms]

Thanks for both of those pieces of information. The better question is that would it hurt anything to make C22/C25 (or parallel the existing 1.2 nF) 100nF? I would guess not. This might be a good piece of advice for those who are building the units w/o a spectrum analyzer "just in case". The noise only looks to be about 30-32 dB down which is enough to want one to mitigate it.

Thanks for the LMH6658MA suggestion. I'm making a list of parts and suppliers for this in North America (or US specifically) for the common good (I'm set -- much of my parts were junk box) on the project. Most of the parts are (far) cheapest from Mouser and the AD868 was one of the parts they do not carry because they do not carry Analog Devices parts. I had a AD868 in the junk box from surplus (we used to have a surplus store at my company and we use a lot of AD parts-- that's super common here). But you are correct in that the National/TI part is 1/3 of the price.

Thanks and cheers.. Fred W0FMS

[g8ajn]

Thanks for the tip, Rob. The noise is apparently coming from the TLE regulators, although on my Spectrum Analyser the noise readings appear to be much lower than those you are showing on your SA. Perhaps there may be manufacturers tolerances involved. It seems a wise precaution to include it so I have modified the circuit diagram and parts list accordingly. It may be too late to make the CQTV publication, but it is easy to piggyback a 0.1 across the existing 1n2pF.
Cheers
Dave G8AJN

[AndyCN]

Not sure if this is totally related - but having just finished my build - and carefully powered up via a current limited supply ( at 12Volts ) I was a little surprised at the 7805 being hot and the board drawing more current than expected. ( Even if I say so myself - I'm a pretty careful builder - so I wasn't expecting it ).

The board was drawing >600mA in peaks !!! ( and it seems RF was feeding back to my PSU , as it was registering volts above the set value , and also volts on an unused output )

It seems that the 7808 I have ( an ST ) was oscillating. ( Is that why I see finished boards using a heatsink ??? )

Adding a 100uF across the 7808 output solved this for me , and the 7808 is now barely warm to the touch ( no heatsink ).

After adding the cap. I'm now only drawing around 150mA

Perhaps 0.1uF on the output and nothing on the input is a bit optimistic ?

[M0DTS]

Hi Andy,

I had a similar problem too and it depends on the power supply used, I added some caps to the 8v regulator like you did to fix it.

I'm seeing ~250mA current from [ Digilite board + VCO + another MMIC amp after the mar-6 ] so your 150mA looks reasonable.

Rob

[w0fms]

Thanks for pointing that out. I also added a 100 uF cap on my input. There isn't a datasheet in the world for linear requlators that doesn't recommend at least >= 10 uF on both input and output. I'm debating on adding 10uF on the output of the 7810/input of the 7805 but if it needed it it would be oscillating there already! That was an interesting omission. I usually use 22uF-100uF for both input and output depending on the unregulated source. 22uF would be appropriate but I didn't have and chip tantalums at that value!

I had no issue but I've been testing with an old home made linear regulator.. so there was 100uF on the output of a 7812 driving the 7810 on my board. But yep, that was a fairly critical mistake.

Nice catch. I also added an extra 0.1uF on the MMIC supply decoupling too.. (I always do .01 and 0.1 on MMICs so the decoupling is over a broad frequency).

73, Fred W0FMS

[M0DTS]

While building another board (v5.7) for Barry G4ZCN i thought i'd have another play with reducing the noise we have seen..
I've now got it pretty low now with even more decoupling, not quite right approach but it does the job!

Add 100nF directly at U3 from ground to output
Add 100nF across C12
Add an electrolytic of a few uF or more directly across Pin 20 to ground on IC1

The latter has the most effect so maybe having isolated Digital/Analogue grounds would be useful in the future.

Noise is now about 55dB down from Tx Level.

73

Rob
M0DTS

[g4bbh]

I'm surprised you needed an extra cap to U3 but not U4. Suggests some inductance in the track from U3 or variations in noise between samples of TLE2426.
I notice that there are not many vias between top and bottom grounds which could also be detrimental to noise.
By the way, it's not easy to see at 50MHz/ but looks like your sidebands are about 26dB down. Is that the best you have been able to get?
Dick G4BBH

[M0DTS]

An extra cap on U4 does not help at all no.

The -26dB sidebands there are part of the 'bell' shape spectrum using the lumped component filter that is used on the Digilite, the real bits you need to be worried about are the sidebands outside these, you can just see them on my last image at ~55dB down but very close in.
On a system with a digital filtering like SR-Systems or AGAF you would not see the bell shape.

The noise i'm relating to is a 'hump' normally +/-50 to 100MHz from the main signal, hence the wide span, you can still see the noise 'humps' but they are a good 20dB down now from what i see with a standard build.

Rob

[g4bbh]

Aah! Thanks Rob. That answers some questions I had about the shape of the spectrum. I had wondered about the 'bell' shape. I can imagine that a digital filter would be desirable but complex and expensive.
Dick G4BBH