AD831 mixer query

[Wendymott]

Hi peeps.
At present I am trying to develop a dual band 2 m / 6 m receiver that will be eventually a RX/TX.
I decided to use the AD831 mixer as it claims to have no insertion loss and a low L / O level requirement.
I attach the circuit of the Mixer, 2 X crystal filter from a FT290 then a BF998 amp into the SSB Filter, again ex FT290. The measuring point is the output of X3. The input is as on the schematic.
The query is this, The output level at the filter for an RF input of -73dbuv is 73 db at 145 Mhz. Ok so far. However the RF output of the I.F with the RF input at 51 Mhz is 6 db down. "Same input level"... I found I was 6 db down, I increased the RF input by 6 db to get the same amplitude out. Looking at the data sheet of the AD831, it is very flat to well above what I want. I tried adjusting the input coupling cap to 1 nF and 100 nF with no appreciable difference. I had noted that 6 M was less sensitive with the previous version of the mixer / IF, but I didnt realise it was the mixer / IF causing the error. Your comments would be greatly appreciated, but dont include any maths.... not my subject.
The L/O is from the Anritsu MG3642 and the RF source is the Marconi 2023.

[Cruisin Marine]

You need to find out where the loss is, is it down at the input to the filter, or down on g1 of the BF Dual gate FET?
Or is it the FET losing 6dB?
Strange it is 6dB, is the LO signal level the same, have you measured it?

Also, Is the filter terminated properly with the correct load and source values?

I don't know those AD chips BTW.

[Julesomega]

The AD831 is a logarithmic amp, not a mixer

[frsimen]

The log amp is the AD8307, the AD831 is a low distortion mixer.

I think you need to have another look at the matching to the crystal filters XF1 and XF2.

In the Yaesu circuit, they are fed from an impedance of about 2.2k and have a load resistor of 2.2k. You are feeding them from an impedance of 133 ohms with a load resistor (R8) of 100k. Remove R6 and insert a 2k resistor between the filter and C12 to sort out the source end. Removing R7 and changing R8 to 2.2k should make the load impedance look a lot better. These changes will improve the passband characteristics of the filter.

It may be that you will need to rework the matching for XF3 too, if it proves to have a poor response.

Why the gain is so different from 2m to 6m, I can't explain. With the circuit you have adopted, the mixer should have some gain, according to the datasheet.

Paula

[Wendymott]

Hi peeps. Cruisin. All levels have been checked, both L/O and RF. All identical. Why would termination of the filters matter that much. The I.F is 10.81 Mhz. The L/O and RF input are set at 10.81 Mhz. Jumping over the X3 to prior the BF998, it is the same result. Paula. I will do as you suggest tomorrow. Its worth a shot. Your last sentence is what this whole saga is about.

[Julesomega]

Sorry Wendy, I was thinking of the AD8310

[Radio Wrangler]

Originally Posted by Wendymott

Why would termination of the filters matter that much.

It matters a lot.

Firstly, the crystal filters are one implementation of a ladder of resonators. The coupling factors between resonators are set by the spacing of the electrodes of the dual resonators, and also by components in the link between crystal units.

The loads imposed on the input and on the output set the Q of these resonators. The other resonators have their Qs also set by the loads, but seen through the coupling factors.

So the loads set the Q's of all filter resonators and, interacting with the coupling factors set the shape and width of the filter. Get the load wrong and the filter gets very ripply before you even notice the width.

Secondly, the mixer chip will have a Thevenin equivalent source impedance. Ideally you want to present something that makes the best of the available power and transforms it to the impedance the crystal filter presents. If you do't do this well, you lose lots of signal. There is need for compromise because you want also to make the crystal filter see the impedance of its dreams, else it sulks!

So you usually try to keep the filter happy while doing the best you can on the loss business.

David

[Wendymott]

Thanks David. Legs duly slapped .
Changes made, yes a difference to the filter response but no change to the difference in RF level.
Still 6 db down @ 50 Mhz. I think I will adjust the gain in the front end to compensate.
At the moment the 2 metre gain is way too high. so I have enough leeway.

[frsimen]

I've had another quick look at your circuit and I noticed that you haven't got a 50 ohm termination on the input to the mixer IC. Without a termination, if you have more than a few centimetres of coax cable between your signal source and the mixer, the voltage at the mixer input will be frequency and cable length sensitive. That could well explain the variation in output level that you are seeing between 6m and 2m.

A 50 ohm resistor across the input to the mixer would be a reasonable start, perhaps a 3dB or 6dB attenuator stage would give better matching.

Paula

[Wendymott]

Hi Paula. Yes you are right.. I will attend to that tomorrow. and come back with a result. At present I have just laid out the pcb for the AD 603 dual chip AGC amp. See if its better than my concoction.

[G0HZU_JMR]

I think it's worth mentioning that the source impedance of the single-ended IF amplifier in the AD831 will be close to zero due to the feedback resistors. It won't be up at 120R as the feedback will make it mimic a voltage source. This definitely means there needs to be a series resistor at its output.

I think the two 82p caps should be something like 470pF or 560pF as this value suits a 10.8MHz IF. This forms a lowpass filter and the aim is to filter away the image frequency and prevent it from reaching the final IF amplifier stage within the AD831.

If the Marconi 2023 is healthy and the 50R coax cable is healthy then the source impedance at the far end of the cable should be 50R regardless of its length. This assumes you have set the 2023 to a reasonably small signal level. The output attenuator of the 2023 should then ensure the source impedance of the 2023 is close to 50R and the 50R cable will also look like a 50R source at the far (AD831) end of the cable. If adjusting the cable length does make a significant difference, then I think something will be wrong somewhere in the hardware setup.

[Wendymott]

Hi peeps AGAIN And the winners prize goes to.... Taaddaa da daa.. Paula. Terminating the RF input at 50 R Levels the inputs to within 1 db.
Geremy the RF level input 1s 100uV. I have retro fitted the 82 pf with 470pf. I already have a series resistor on the output of the mixer to 2K2.
I Can now move onwards a bit. Thanks for your comments.

[G0HZU_JMR]

Hi Wendy. Glad you got it working OK. I can't really explain why terminating the AD831 end of the cable made a difference unless there's something odd about your sig gen output impedance.

If the source VSWR of the sig gen is as poor as 2:1 then you can definitely expect to see up to about a 6dB variation in voltage at the far end of the cable depending on cable length. However, I would expect your 2023 to have a VSWR very close to 1:1 at 100uV EMF because of all the attenuation in the 50R output step attenuator of your sig gen.

Is there any way your sig gen attenuator could be faulty in some way that affects the source impedance of the generator? Could there be a dodgy RF adaptor with a poor VSWR at the sig gen end of the cable?

Note that placing a 50R termination across the AD831 input will degrade the noise figure of the mixer quite a lot so I'm not sure this is a good solution unless you plan to put lots of preamp gain ahead of it.

[G0HZU_JMR]

I do think it's worth checking the quality of the RF cable and maybe also your sig gen if you are seeing a 6dB variation. I know you don't like maths but there are calculator programs that can predict the amount of ripple to expect, and the ripple depends on the source VSWR of the sig gen and the VSWR of the load and the cable length.

The program below predicts a ripple of +3dB and -2.2dB (depending on cable length) if the sig gen source VSWR is as poor as 2:1. I've assumed the input VSWR of the AD831 is about 14:1 at VHF and this is why I've chosen 14:1 as the load VSWR.

If the source VSWR improves to 1.1:1 then the mismatch ripple improves to just +/- 0.35dB. However, I'd expect the source VSWR of the 2023 to be even lower than that at 145MHz when set to 100uV output. That's why it might be worth checking your cables and connections from the sig gen to see if something is causing a mismatch problem. Are you sure you are using a 50 ohm coax cable for example?

[G0HZU_JMR]

To show what I mean, I just tried measuring the source VSWR of my Marconi 2024 sig gen when set to 100uV emf at 145MHz and it measured about 1.04:1 across HF and through VHF. By 1GHz it had degraded to just under 1.1:1.

I also tried feeding the output to a high impedance scope input set to 1Meg input and swept the sig gen across a few MHz to about 200MHz. I used a 2m long 50R coax cable for the connection. The output emf at the end of the cable was quite constant at every frequency. Any ripple was quite subtle. I'd expect your 2023 to be the same. I had to use the FFT function to see the tiny 100uV signal. I also did the test at about -35dBm output so I could see it on a regular scope voltage vs time display and the result was the same. hardly any visible ripple.

At high sig gen output levels, the source match will degrade because there will be less output attenuation in the sig gen and the ripple seen on the scope will get much worse. A lot depends on the make/model of sig gen. My 2024 performs badly in tests like this at high output power levels as it is the high power (+25dBm) variant.

[Wendymott]

Hi Jeremy. The coax lead is RG58 from the 2023, about 3 metres to a diecast box with BNC connector. Then a BNC to SMB adaptor with RG174 cable (1 m) to the DUT. The RG174 is ex rally "pulls" from who knows where, but properly terminated. The mixer input would be connected to the "front end" see attached. At present the gain for each input is way too high..
it will be trimmed when I get the I F stage sorted.
I am wondering if I am being too pedantic. The gain of the 6M strip can be matched later to equal the 2M strip. I do know the gain of the front end should not be excessive, and that the I.F should do the heavy lifting.
Now... where did I put that University degree ?? Hmmmmm never had one.

[Radio Wrangler]

Originally Posted by Wendymott

Now... where did I put that University degree ?? Hmmmmm never had one.

If you look at almost all the people who are quite capable in their fields in science and engineering, then look at all the notes from their lectures while at university (if they went), You'll find that very many of them are operating on knowledge acquired after they acquired the fancy paperwork. Also, once you get to a certain level, universities start giving out freebie degrees. These things ought to be evidence of the true value of degrees. Hell, they even give freebies to politicians!

("Price is what you paid, Value is what you got" Warren Buffet)

Once you've had the job of bringing a fresh graduate up to speed in their first job in engineering you see things in proportion.

Universities teach people how to perform analysis, in the main
Employers pay people to perform synthesis, in the main.

So you need something beyond degree level in order to be much use.

Don't worry about the things. They are a distraction. Unfortunately personnel departments can't look beyond paperwork. But that's their company's loss. A company makes more money out of an employee than the employee makes out of the company, so if the company misses the best guy, theirs is the greatest loss.

Get your bum on an interview seat and you'll get grilled on your post grad jobj, your project and get some problems to solve. You won't get asked about lectures and exams. Effective interviewers have to know what really counts.

David

[Wendymott]

I think I was being "ironic" David . When I worked at Bradford Uni in Physics electronics workshop we had more idea about putting projects into reality than any academic. Thats what we were there for. I was told early on that Students were there to learn how to think, not do. "Some" PHD students were good though. No I will stick to "bumbling along"