More experiments with SiC FETs on 40 meters - different devices this time

[steve_qix]

I am not satisfied with the performance of the 40 meter RF amp I built for Rattlesnake Island (4 of the larger Cree SiC MOSFETs in class E), so I am going to try something different:

Using 8 of the new 900V devices in this amplifier - two in parallel per module with individual drivers for each.  Two of the 900V devices will actually be capable of more power (higher current), lower R R-S on, be easier to drive (approx 650pF of gate C) and two of the 900V devices cost less than one of the C2M0040120D devices I am currently using.

I suppose the major disadvantage to using the smaller devices is that the amp is harder to build - more drilling because there are more devices.

I've been having driver failures and this is not good.  It may be that the IXDD414 / IXDD614 drivers can't look into that much C at 40 meters.  I think I'm at the upper limit of the frequency range of the drivers anyway, so the lower capacitance should make them a lot happier!

I will report back!!

Regards,  Steve

[steve_qix]

Ok, I finally got the RF deck together - was very busy in there doing everything but ham radio!!

Anyway, so far so good.  Got it up to 600 watts so far (going slow).  The RF bypass caps are heating up, so I'll need to add some more caps in parallel to handle the current.

Otherwise, linearity is excellent and the heat sink is COLD.  This one, so far, is more efficient than the last one.

Uses two of the 900V devices in parallel per module; 4 modules for a total of 8 devices.  Each device has its own driver.

More reports will be forthcoming as I get the power up to full.

Will most likely be on 40 meters this evening, assuming everything is functioning properly.

Initial tests are encouraging, and I've had 2 QSOs on 40, just a little while ago (it's 5:00 E.S.T. now)

[W3GMS]

Thanks for the update Steve!  All makes sense and easily understood why its working much better now as compared to the earlier configuration. 

Joe-GMS

[ka1tdq]

You were heard over here in Arizona.  I got on later, but someone asked me if I knew you (1 land and all), because they had heard you on.  Actually, the station was an XE2/W7 station, so he actually heard you in Mexico. 

That's pretty good, because the band is fairly noisy over here tonight.

Jon

[N2DTS]

I was wondering if you could be heard on 40 meters with such low power (600 watts carrier).

[steve_qix]

You were heard over here in Arizona.  I got on later, but someone asked me if I knew you (1 land and all), because they had heard you on.  Actually, the station was an XE2/W7 station, so he actually heard you in Mexico. 

That's pretty good, because the band is fairly noisy over here tonight.

Jon

That's pretty good distance for summer!  At some point, maybe we'll actually able to connect !  The static was fairly intense that night.

Nigel and I had quite a nice chat about output transformers and the like.  I think mine are not working nearly as well as they should - and am going to use a different design and core material.  Cores are on order, so when they get here, I'm going to do some experiments.

Regards,  Steve

[VE3ELQ]

You were heard over here in Arizona.  I got on later, but someone asked me if I knew you (1 land and all), because they had heard you on.  Actually, the station was an XE2/W7 station, so he actually heard you in Mexico. 

That's pretty good, because the band is fairly noisy over here tonight.

Jon

Nigel and I had quite a nice chat about output transformers and the like.  I think mine are not working nearly as well as they should - and am going to use a different design and core material.  Cores are on order, so when they get here, I'm going to do some experiments.

Regards,  Steve

Steve you were 40db over S9 on 40M and Hi Fi great so its certainly working.  But as we discussed I think you will find a toroid O/P xfmr a little better once you fiddle with it and get it optimized for 40M. LF to hearing how it works out.
73s  Nigel

[steve_qix]

Well, the new transformers are _MUCH_ cooler than the former ones.

Have not tried the toroidal style yet - first just wanted to try different material.  In this case, type 61 instead of 43.  The 61 runs much cooler than the 43, and everything looks good.

Still having RF bypass cap heating.  That's the next thing to do.  Each module does pull 5.6 A, so the RF bypasses have to be able to withstand that much current at 7 mHz.

[W4AMV]

I have been looking at 2 Cree SiC FETs, the C2M1000170D and the 280120D. The 170D has a lower Ciss at 190 pF and considering simple push pull class C. There are models for the devices as indicated in prior post. Not sure how applicable (accurate) to our applications, although most of the key parameters are there. In LT Spice, it is easy to get S parameters and from that stability. What I have noticed, as the Vdd supply goes up, the stability is horrific. From say 28 V -to- 60 V you really need to do something to tame this device. The easy solution found is to add series gate resistance in the RF path. A 120 ohm series R allows the device to be easily matched and the device gain is still north of 20 dB on 40 meters. Not sure if the application of some series R in the class E designs would be useful to address the failure mechanism.

Alan

[steve_qix]

I have to say, I have not personally run into any stability issues with these devices.

I'm just wondering... the gate Xc is approximately the same (at 40 meters) as the proposed 120 ohm series resistor.  This seems as if it would significantly reduce the gate drive to the point where the device may not be saturated (did not look at the curves for the device to verify, but other devices in the family need a good amount of gate voltage to achieve saturation) and therefore the efficiency significantly compromised.

Unless you were thinking of using a high drive voltage ahead of the resistor, and this has its risks - if the gate, for whatever reason is kept turned on long enough, eventually  the full driving voltage will be reached.

Please expand on your thoughts - very interested.

[W4AMV]

Hi Steve, thanks for feedback. I agree, a series R of that magnitude seems large. And I have to emphasize, this is from small signal s parameters extracted from the Cree model for that device. However, the next step is to validate if measurements hold up against simulation. The addition of series R is typical in large GaN devices particularly at low frequencies where they have significant low frequency gain. However, the R is typically 15-30 ohms for a fairly large device. My initial idea was to simply match a single device biased at just near the threshold V, about 4.5 V and Vdd of 50V. Again, I would add the series R to begin as the simulation indicated horrible instability. As Vdd is lowered, the device QUICKLY becomes tame. Of course the insertion gain drops like a rock. Finally, I need to make a note as to the value of S11 from the simulation and see if that and the remaining S data make sense when a 120 ohm R is inserted. It was a very nice easy match and the gain was very respectable.

[W4AMV]

Attached is the S data simulation copy shots from the simulator. At 50 V, the S data demonstrates a quite unstable device while at 24 V, the device is quite well behaved. The S data makes sense as it tracks the device performance, increased gain and greater isolation and so on as the Vdd is raised. Drop in gain, improved stability as Vdd is lowered. All the S data values are numeric, not dB. Yes, the Cin is a bit large, but the Rin is also larger than I would have thought. If the Cin is absorbed into the match Steve, then the series R near 120 ohms would make sense. You need to tune out the reactance. Of course all this needs to be validated.

[VE3ELQ]

It appears that you have modeled the FET with bias placing it in a linear mode.  Was that your intent??  These are switching FETs which we generally run at zero bias switch mode class D or E which is what they are designed for and is very different.  They make very poor linear amps. There are LDMOS FETS much better suited for linear applications.
http://www.nxp.com/documents/data_sheet/BLF188XR_BLF188XRS.pdf

[W4AMV]

Hi Nigel, yes, that was my intent. I agree they are certainly very suited for Class D,E, however, there is no reason that they could not be operated Class C,  for example as I can control the conduction angle via the bias. Of course, the S data is not straight forward to obtain for the switching mode and I am assuming that I can obtain S data from the Cree model. The devices were placed in the linear mode, here, just to obtain some S data. I emphasize, an assumption on the use of the Cree model for this application. In any case, I made an error in obtaining the S data from the model as I have the 50 ohm load on the drain side in series with the drain. I have effected the drain voltage and consequently the S data. Indeed, the device at least over the short range of frequencies at HF IS stable. However, I show a series gate R anyway, to provide a match and toss away some gain! It is not needed and proof of that requires building the circuit. What I have for simple evaluation is shown attached.

[W4AMV]

Adding the corrections to the prior post here. The S data from LT Spice corrected and the match added here. The particular device used for this investigation was in hand and agree that Class D,E is appropriate. However, for those who would like to apply to other Classes of operation or consider push pull, wide band, and so, perhaps these devices have a place as well. It is nice to be able to build the match in LT Spice and validate operation. Transient analysis could be used for power amplifier and push pull investigation. The device is more stable than first indicated from prior simulation, again set up error in the simulator. As seen in the second picture, significant gain and quite good match at 7 MHz with a simple set of components.

Alan

[WD5JKO]

The link below discusses the technology drivers behind the new SIC technology. From what I can tell, CREE is a small player in this market. The company I work for is being approached by multiple companies to make an ion implant machine custom to this emerging technology. In this case, a silicon carbide wafer is heated to ~ 700 degrees Celsius and then implanted by high energy ions from aluminum and other substances. Without ion implant, the method used is an older diffusion process done at much higher temperatures.

http://electricvehicle.ieee.org/2014/10/31/market-analysis-wideband-gap-devices-car-power-electronics/

Jim
Wd5JKO