Dual gate MOSFETs Equivalents?

[Wendymott]

Hi Peeps. I am conducting a blatant "copy" of the FT290 ...... well not a copy as such... but a rebuild using more modern components, such as SMD devices.
The rescued inductors are too small for my needs so I am using Toko 10mm form types. The problem I have is..the 3SK59 (RF Amp), 3SK51 (Mixer) and 3SK73, I.F amp all have the same parameters in the "Allradio " semiconductor listings. I want to use the BF998, the pcb is laid out and made using this device in replacement for all 3 devices....... I know this may not be acceptable, but I am unsure why. I have made a "fair" copy of the FM section, with a couple of "birdies" eminating from the RF front end, I am now approaching the SSB section as in MC1459...it has two pre IF stages using the 3SK73, should I use the salvaged devices or continue with the BF998's. See attached SCh.
The front end is the Mutek module, now built onto the pcb, rather than being "out board", then the rest is FT290, but RF chokes in the drain of Q2 Q3 and Q4, as they seemed to give more gain... maybe thats a no no.... The copper is double sided, topside being ground plane

[Radio Wrangler]

You have a closing window of opportunity. Dual gate MOSFETs are going away, part number after part number. There seem to be no big users left, so the semiconductor manufacturers are responding. I got bitten because I'd used some new-ish dual gate devices intended for low supply voltage, using integral current mirror bias controllers. Quite smart, but now unobtainium. The plain old high voltage dual gate parts are going obsolete more slowly, but still moving.

By the way, the Japanese makers put a lot of work into fiddling their layouts to get things stable and reasonably clean despite hacked-about ground areas. You may have to redo all this if you run into trouble.

I'd probably go multilayer and get solid planes in, but that's me.

David

What is the difference twixt them and a discrete component cascode? Apart from the obvious reduction in component count. Perhaps a FET in the bottom and bipolar at the top.

[Cruisin Marine]

I have made a "fair" copy of the FM section, with a couple of "birdies" eminating from the RF front end,

Be very aware of self resonances with RF Chokes in the drain, especially with nicer, perhaps higher gain devices. I have more often than not had to tame those chokes with a parallel resistor to get them stable.

I see there is a resistor on Q1 drain cct which has no parallel C, but that may need decreasing in value to stop it going a bit nuts.

[G6Tanuki]

This is a kinda parallel of my request for advice on 'modern' U310 FETs.

The classic RCA 40673, European BF981 and Japanese 3SK88-type dual-gate MOSFETs are no longer available from reputable/properly-heritage-traceable sources.There are perhaps some SMD-packaged equivalents which can be fitted to a little square of PCB with pin-outs that natch the old-style dual-gate MOSFET packaging.

In some cases these will give rather-better performance [gain, noise-figure, IMD] than the devices they are substituting for.

[Wendymott]

Hi Crusin (Andy). The front end is a direct copy of the Mutek preamp, thus I saw no reason to "fiddle" with it. Fortunately I had a spare, sent by a kind member of the group.
Ok David...but no one has answered the question... how do I equate a BF998 to the other devices.. do I just ignore the parameters and do it anyway... Obviously Yaesu chose those devices for their parameters originally... the 3SK59 was chosen for the front end..was it a "noise" factor ?, as it has AGC on it via G2, is it better than a 3SK51 for AGC action ?
I guess "Tanuki" that its just a case of suck it n see. EDIT MC1459 should read MC1469..

[Cruisin Marine]

Hi Wendy,
I dunno if you have seen this, it may interest you as it specifically refers to the BF998. You probably know all this anyway, but worth sharing.
You would need to compare devices via data sheets of course. They are all similar though, but very "gainy" as they are cascode devices as you know.
I guess the answer to your question is to literally compare results with different FET's and boards and trim values to match if you wish to get close results (but do you, or perhaps improve on the original?).
As per AGC results you would need to plot results on graphs and overlay those to see the difference.
Apologies if I am stating the obvious.

http://w7zoi.net/lna50.pdf

[G0HZU_JMR]

Ok David...but no one has answered the question... how do I equate a BF998 to the other devices..

I think you will end up with lots more (unwelcome?) gain in each BF998 amplifier stage. The BF998 is a DG MOSFET with really high gm. It's something like 24mmho. This means that unless you find a way to reduce gm (by biasing it softly) or you load down the drain then this device will provide lots of RF gain at 144MHz in a preamp.

By comparison, the old BF981 has a gm of about 15mmho on the datasheet. I do have some s-parameter data for the BF981 and the GMAX at 144MHz is about 29dB at 10mA Id. At 50MHz it is about 33dB and at 12MHz it is about 39dB. These gain numbers for the BF981 are already 'big' but the GMAX numbers for the BF998 are probably going to be even higher.

I had a quick look over your circuit and it does look a bit strange in places. I'm not sure how much image rejection it will provide. I think the original 290R had a (narrow) varactor tuned image reject filter at the front end.

There also seems to be a lot of RF gain ahead of the first mixer with those two BF998 stages ahead of a MOSFET mixer. This will make it very prone to overload from strong signals and the input IP3 could easily be as low as -30dBm but that is just an initial guess. Your circuit probably doesn't do much to suppress image noise (from the second BF998) at the first mixer and also the first IF filter doesn't look to be terminated the same as the 290R circuit terminations at 10.8MHz.

A classic front end with a BF981 followed by an image BPF and a diode mixer and then a low noise IF amp would manage an input IP3 in the ballpark of 0dBm. The overall system noise figure might end up being about 4dB if (diode?) TR switch losses are included. This would be a reasonable performance benchmark. This would be easy to improve on but would be at least as good as most 2m radios of the last 40 years or so.

I also read through the W7ZOI preamp article and the design choice of the BF998 seems a bit strange here too. Who would really need a preamp with 28dB gain at 50MHz? This gain figure seems unrealistically high for a typical 6m application.

[Oldtestgear]

Try Linear Integrated Systems in Fremont California. They are still manufacturing a wide range of JFETs & MOSFETs. If they have want you need then the fun starts trying to buy the parts. I have used them in the past & the minimum order charge (via a UK disty.) was reasonable.

Hope this is of help.

Phil

[Wendymott]

Hi Jeremy. THANK YOU for the circuit analysis. In the 290, original, as you may know, it has a DG Fet with a bandpass filter, with a varicap "tweeking" circuit, where I think the voltage is derived from a part of the circuit I have not used, thus I decided to add the Mutek preamp, onboard, I also found that I was not getting the same sensitivity with just one stage of IF gain, so I added another stage pre Xtal filter, I was going by the input to the FM Detector chip pin 16, and comparing it with the working FT290. I am still about 5 db down. I will refit the Tunable filters in the drains of Q2,3 and 4, and possibly damp them if it gets too "twitchy".
I guess enthusiasm is no match for ability, some of us are academic some are not, but we get by..... still I have something to "get up" for.

[Wendymott]

Sorry Andy, I missed your comments,, I zoomed in on Jeremy and didnt see it. See my comments to Jeremy "above". And to Phil. yours came in while I was typing my reply to Jeremy. I think I will stick to the BF998 and get out a "Big stick" to tame them

[Cruisin Marine]

Sorry Andy, I missed your comments,, I zoomed in on Jeremy and didnt see it. See my comments to Jeremy "above". And to Phil. yours came in while I was typing my reply to Jeremy. I think I will stick to the BF998 and get out a "Big stick" to tame them

No worries Wendy,
Here is an excellent little article which I will share, it gives some great little tips and goes in to things in detail, which I am too darn lazy to do https://www.qsl.net/sv1bsx/vhfpre/vhf-pre.html

It has a noise figure of less than 1dB, so overall a winner.

[Wendymott]

Hi Andy. Interesting, but limited to a certain portion of the band. As with the variable caps, the nice thing about the Mutek is the bandpass, very controllable. Looking at the FT290 again..I think the varicap diodes, may just shift the tuning in "lumps" as in..SSB lower end of 2m FM upper end of 2m, this would allow, increased gain, for each mode, without sacrificing too much loss. I may just have a fiddle with that and see where it goes.

[Cruisin Marine]

Interesting, but limited to a certain portion of the band. As with the variable caps, the nice thing about the Mutek is the bandpass, very controllable. Looking at the FT290 again..I think the varicap diodes,

Sorry Wendy, it wasn't meant to be a final design for you to use, more a link to drive you in the right direction to achieve your design goals. To fire your imagination if you like.

It is a fine design, with coupling and stability at heart and even an attenuator to load the o/p properly to ensure even more stability.

There are some good tips and guidance in the article, even though it uses trimmer capacitors, there is no reason whatsoever it in itself could not be modified to use varicap diodes with minor mod's, but the intention was more to try and answer some of your questions about design and devices and how to accommodate different dual gate FET's. therein, or drive you towards that, and not intend to suggest you in any way use the exact design yourself.
Without me having to re-write or re-word the ideas in the article, I thought it was good to pass on.
cheers

[G0HZU_JMR]

If it helps, a while back I measured a BF981 on my VNA and generated a full s2p data file and this also included the noise parameters based on the datasheet. This enabled me to model the complete Mutek preamp circuit that I believe is the same version fitted to my FT290R mk1.

The Mutek design really is rather elegant. Other online designs I see are anything but elegant in comparison. There are a few basic design rules with JFET and MOSFET preamps and I learned these the hard way many years ago.

In terms of noise figure, the Mutek design seems to have close to the optimum LC input network that achieves low noise figure. Also, the drain loading looks to be what I would expect (for sensible signal handling) and it also manages to provide an adjustable RF gain and it also achieves some useful additional selectivity in the bandpass filter. I think it has been designed quite carefully to achieve low noise figure, good stability and a useful improvement in out of band rejection. The simulation predicted the Mutek board has a 1.2dB noise figure at 145MHz. This assumed a sensible Q for the input inductor.

Sadly my Mutek preamp in my FT290R wasn't always stable but this was later traced to a poor ground connection that introduced an external feedback path. Since this was fixed it has been fine.

I'm not so keen on the design of the other preamp designs listed on here but that is just my opinion.

[G0HZU_JMR]

I also had a rummage through the s-parameter data files bundled with my Genesys simulator and was surprised to find that Genesys includes the BF998 in amongst the data files in the Infineon subfolder. It also includes noise data in some of the s2p files for the BF998.

There are s2p files for the BF998 for (Vds, Id) (8V, 5mA), (8V, 10mA) (5V, 10mA) and (10V, 15mA).

I can post these files up on here if they are of any interest? With these files you could simulate the first two stages of your front end for gain, selectivity and noise figure. The noise figure is easy to predict anyway... you will get just over 1dB because you have loads of front end gain.

[G0HZU_JMR]

Alternatively, I could do it for you if that is easier? It would have to wait until the weekend though.

I did try slotting in the Infineon BF998 model into Genesys in place of my BF981 model and the gain of the Mutek preamp went up by just over 3dB.
I used the 8Vds 5mA model for the BF998 as I see you have included a source resistor. There isn't an Infineon model for 2mA or 3mA Id. If your Id current is somewhere around 2mA to 3mA you might end up with the same gain as the BF981 biased at 8-10mA with no source resistor.

Genesys predicted the noise figure was essentially the same at about 1.25dB when the BF998 model was put into the Mutek preamp.

[G0HZU_JMR]

I had a look at a basic design spreadsheet this morning for a typical 2m front end using a DG MOSFET for the first mixer. I'm rusty with mixers like this but it should be possible to achieve the following with a mixer like this at 144MHz.

10dB mixer noise figure
10dB mixer gain
10dBm mixer output IP3 (OIP3)

If I put these numbers into a spreadsheet and include a BF998 preamp with 23dB gain and 1dB noise figure (with OIP3 15dBm) followed by a 3dB attenuator followed by an image reject BPF with 6dB loss followed by the above mixer and then a crystal filter with 4dB loss and then an IF amplifier with 4dB noise figure you should end up with a system noise figure of 2.5dB and an input intercept point of about -14dBm. If you spent time polishing the mixer design and the loss of the image reject BPF the noise figure might improve to about 1.8dB but I'm not sure it is worth the effort.

By the time a TR switch is included and also an input LPF for the transmitter the system noise figure may end up at 4 or 5dB. These numbers are OK and probably similar to most 1980s 2m mobile radios that use a similar front end design.

Can I ask what are you using for the 134MHz local oscillator for your testing? I don't think you can use a regular lab sig gen (too noisy).

Also, your mixer interface to the crystal filter looks strange as I mentioned before. Normally the mixer output pin (drain) should be presented with a low reactance (close to a short circuit) at the RF and LO frequencies. If you look at the output of the mixer on the FT290R circuit there is a low impedance path at VHF for the RF and LO (via C20 and C21) and there is some resistive damping (defining?) at the input to the crystal filter via R11.

The resistive damping prevents the mixer seeing a huge impedance at the crystal filter stopband frequencies. This will help with signal handling for local signals that are a few channels away. Your circuit doesn't have any of this and this is unusual and I'm not sure what effect this will have on the mixer performance.

[Wendymott]

Sorry peeps, not been monitoring this thread..as I should have done... had other stuff pressing...Well Jeremy...you have certainly given me something to think about. I replaced the mixer Drain inductor "choke" with a tuned inductor, and also one in the post I.F amp. As it was only completed yesterday, I decided to test today "sat".
The L/O is a DDS working at 156 mhz ish. I have also tested at 134 ish and there seemed to be no real difference. The I.F is 10.81 Mhz, parts recovered from two 290's.
As I have said before, there are lots more technically competent members than me / I in the group. I have attached the update sch.

[G0HZU_JMR]

Thanks for the update Wendy. I'll have to think about the DDS LO. Normally this type of LO would be fine but the gate 2 to drain isolation of a DG MOSFET mixer won't be the same as the LO port to IF port isolation of a typical double balanced (diode) mixer. Therefore, the wideband noise floor of the DDS local oscillator might be able to creep into the signal path and make the receiver appear quite a bit deafer than it should be. I need some time to think about this...

A regular LC based VCO (as used in the FT290R) will probably manage a far out noise floor approaching -170dBc/Hz. A DDS might be 10-15dB worse than this. It could be much worse if the internal clock multiplier is used to derive the high speed DDS clock from (say) a 25MHz crystal oscillator.

Some lab sig gens are also really poor in terms of ultimate noise floor. This usually isn't a problem when driving the LO port of a double balanced mixer but I think the noise floor issue becomes much more critical with other mixer types especially if the LO port requires a large amplitude signal to maintain conversion efficiency. The absolute power of the LO noise in dBm/Hz might be quite high in this case.

This issue can be fixed with a narrow BPF on the LO output so it's no big deal really if the LO noise floor is a bit high.