The 6M-160M HPSDR Transceiver Project Begins

[WA1GFZ]

Tom,
Do you have another computer that you could load PSDR into?
It sounds like you are not saving set up information. Let me look at my set up to see if there is a save button that needs to be pressed for RX. Rob showed me the one in the TX page for audio settings. Go to the main HPSDR page where you install mercury and penny with the clock source. See if there is any additional buttons you need to press to save the set up. My Version of PSDR is 1.10 SVN XXXX so goes back a way. I'll give you a call tonight if I find anything. I think we went through all the set up stuff so I'm at a loss. We have the same settings but your P SDR may be newer.

[WA1GFZ]

Tom, I had MacaFeece anti hiv on my shack computer for a while. It drove P SDR crazy. I removed it and don't put the machine on the net. Do you have any anti hiv software in that machine?

[K1JJ]

I have Avast in their and deactivate it sometimes. Trouble is I do use the computer on the web sometimes for HPSDR related stuff right now. Later on I will disable it when the rig is running right.  Yes, that could be some of the cause, no doubt.  

BTW, It appears that Penny's locked up spurious thing happens only when the PA db is set at 38db. When set higher (42DB) it doesn't occur at all. Little by little ironing out the bugs.  Now if Gerd finds the spike issue, I'll be all set for now.


I'm looking at the SS linears here now.   It wud be nice to run the CCI dual amps at say, 40w out. Maybe that wud keep them clean in class A.  I don't need much drive for the Dr Love chain on ssb or AM.


T

[WA1GFZ]

I think the erb amps are cleaner since they have a much better transformer design that will work on 6 meters.. Maybe the CCI amps will be ok at lower voltage say 28 VDC I'm not sure you could get 40 watts with 1 stage of gain.
Erb will easily give you clean 40 watts. Just need to attenuate the input.
I bet you could eliminate the first stage and feed .5 watts PEP into the driver but it would take some transformer design to match the new input. 
I only download HPSDR stuff with the computer when I have to. Otherwise the network cable is disconnected. Metis will be the new interface to that connector.

[WA1GFZ]

Well Tom,
Looks like you set the collective in motion based on all the email traffic.
We have not had a good bug in a while. Watch, one of these guys will solve the problem with the spike quickly. Interesting on the gain thing. Maybe I don't see the problem since my output transformer is backwards

[K1JJ]

I'll have to see if your Metis does anything better with the current boards.

Hey, just took the Erb solid state amp out of moth balls. Using a linear power supply with 0-35V, I found that at about 20V and using 0.3 p-p volts of drive (0 dbm) I am getting out  50W.  I set the bias so that the finals and driver idle warm.  Thats about 1.8 mW of drive, if my calcs are right.    I have 50V peak-peak out into 50 ohms. (50W)

So that's about 43db of gain.    Not bad.  I will definately need an attenuator for the HPSDR which is capable of 400 mW.

Do these figures look right for the ERB?

I do have one MRF-150 that is running hotter than the other. Maybe it's not seated right on the heatsink. If I can get that one cooler, I can run them both up idle some more. But looks like a winner and I will use it.  It looks like Erb will not need a fan. I had it running at 50W out for 30 miinutes and the sink is barely warm.

T

T

[WA1GFZ]

Tom 50 watts is 50 volts RMS so you need to multiply thaty by 2.8 so you really need 140 volts peak to peak output. Run the whole strip ad 24 volts and you will easily get 100 watts PEP. Reducing the power to 50 watts will improve IMD. You will need to bias each MRF150 at 3 amps to get best IMD performance. Look up the motorola spec on the driver to get it class A.
I bet that will be a couple amps. The first stage is biased at class A running at 900 ma so you don't need to mess with it. So you are talking close to 9 amps bias to get best IMD performance. At 24 volts that is 216 watts of heat.
Even my heatsink with 750 square inches and a glob of 1/4 inch copper spreader I think I might blow a little air at it running class A. Those motorola heat sinks do not have enough surface area for that kind of power without blowing air past them. Yup over 40 dB of gain I agree. Running higher bias may increase gain. I was wondering about doing more RF feedback to reduce gain. I have not run any numbers yet though.

[K1JJ]

OK, sounds good. There shud be plenty of headroom cuz I only need 30W or so for the Dr. Love linear chain.


In general, which should give better IMD at maximum heat idle?

1) Running at 18V with higher idle current

2) Running at 24V with lower idle current

I wud idle setup each until they are at max heat.  


  
I will run some IMD tests later to see where the sweet spots are, but I don't think I could get 216 watts of steady state heat from that particular heat sink system. Though the sink is pretty good sized, with two 3/8" aluminum slabs and large fins below that.  But not as big as the CCI amp. I don't have a copper slab in there, so that's the limiting factor to quick heat transfer.

Gonna build a small 50 ohm  input RF attenuator for it today.  50 ohms to gnd, variable pot in series and 50 ohms to gnd.  What value shud the pot be to give about 30db attn max?

T

[WA1GFZ]

Tom,
I don't have any IMD curves for other voltages in front of me. Most MIL rigs ran 28 VDC on the final. I would think the higher voltage would be better because all the Motorola curves are at 50 volts.
The first stage is set up for 24 volts so run it there. 
I don't remember how I set up your bias maybe two pots. My erb I went with 4 pots and a master pot on the voltage regulator. At minimum R for the master I set all the 4 pots for
suggested current at class AB. The the master can increase the bias on all the pots to run them up higher.
Attenuator 200 ohms should get you close maybe even 1 k.
I'm running 16 dB of attenuation pads but your penny output transformer is right so might need a couple dB more.

[K1JJ]

Built up the adjustable input RF pad and got Erb running with HPSDR. Seems to work FB. However, the IMD is not any better than -32db or so. It's about the same as the FT-1000D.  I tried it barefoot, normalized the signal and the HPSDR is way way cleaner than when running the ERB.

I tried various VCC voltages on Erb and different idles. It seems the higher Erb idles, the better the IMD, of course, but I need to run the VCC down around 16V for best prerformance, otherwise it gets too hot.

I keeping playing with it to see what I can do.  But it IS a working system as-is and would pass the normal IMD tests.  I need more like -40db 3rd to reach the minimum goal of the cleaner Dr Love amp chain.

T

[WA1GFZ]

Yup, You need to run the bias up to get better IMD. I don't have any curves for low voltage operation so you are on your own. You need to make heat to make clean, no free lunch. Note as you lower the voltage the bias on the first stage goes down and could come out of class A.
There are two trim pots near the input stage FETs that set their bias. It wants to run at 900 ma both devices. There is a current sense resistor where you can monitor voltage to get current. check schematic.

[K1JJ]

Yes, I see those pots for the predriver.

Now that I have a clean -50db 3rd IMD exciter (HPSDR) I'm going to drive the Dr. Love linear chain with it and see just how clean those are. Before, I only had the -32db IMD FT-1000D exciter.  This will tell me if I am wasting my time with these low level stages or not.

Worst case is I cud build up a class A tube predriver.

T

[K1JJ]

Well, I tested the HPSDR into the Henry amp, which is modifed for 1500V and loaded heavily, giving about -45db 3rd.  It is definately about 10db cleaner than the Erb amp.

I tried more Erb tests but find it is cleanest at lower voltage but more current.  I realize the MRF-150's want to run at 48V and more heat to be cleaner. 

All I need is about 30 CLEAN watts to drive the tube amp.  The Erb seems like a lot of trouble to sink more heat, so I may look into building a simple two stage tube amp with negative RF feedback.  Not sure what to use for tubes at this point, but I've seen the 6146 driven by  a 12BY7? using neg feedback that has worked well.  Or maybe a 6146 driving an 813 in class A?  I dunno.

The Erb wud be good  as-is for low power and normal use, but the Dr. Love chain is cleaner and deserves better.

BTW, I had a couple of Penny lock ups into the amps. I added a 50 ohm resistor across the Penny output and this seems to have stabilized it. I think the spike shock was setting off the instability of a less than perfect load into the amps. As Penny unkeyed, she went crazy w/o the external load.

T

[WA1GFZ]

I told you the attenuator was a pi with two 50 ohm resistors to ground. The pot is the top element series. Both penny ad erb want to see 50 ohms. At 16 volts the first stage may be coming out of class A. Why would you want all that tube BS when a simple BB amp will do the job 160 through 6M with one power supply.
You just need to do the heat sink right which seems a lot easier than designing a tube driver that will be a lot more work. How many man hours do you think it will take to make it work 160 through 6. The tube will need to be close to class A to give you better than -40 dB IP3 so pick your heat source.
A push pull stage is a lot easier to make clean.

[K1JJ]

"I told you the attenuator was a pi with two 50 ohm resistors to ground."

Yes, that's what I built.

Oh I see...  I had the pad in the linear amplifier after the relay so the receiver would not be affected. During unkey it was out of circuit. I should have put it into the HPSDR cabinet right on the output of Penny's RF out so it's always there.   Well, I added the 50 ohm resistor there so we're all set for now.  I may not be using Erb anyway.

I dunno. I think these SS amps need some additional serious negative feedback.  Plus,  running class A is mucho heat sinking - not easy to implement - and maybe tubes are easier to cool.  Even running a small signal thru the SS amp, that can be considered class A, is not all that clean.  In contrast, my Henry is cleaner as well as Dr Love.     Maybe even a simple class A GG 813 driven by a class A 6550 in GG  (or whatever tubes are best for linearity) wud be stable, easier and work well.


T

[K1JJ]

Hey, I just came up with an idea that may work.

The FT-102 -  I have two left here:

"Third order IMD Better than -40dB (14MHz, 100W PEP) "
http://jvgavila.com/ft102_0.htm


Use the linear amp inside. Drop the power down to 30 watts out and it may clean up more. It's rated 160 w out.  I could put that unit on the table and have a smaller footprint than the exisiting SS amp and PS.  And see at least -40db 3rd.  It will accept a 10mW signal IIRC.

It also says:
"Negative feedback level Approx. -6dB at 14MHz "
Heck, why not increase that even more than -6db since there's extra gain available. I did that with a TMC amp once until I could barely drive it... :-)



I'll look into this and see how it works out.

The key here is they are doing this cleanliness with RF negative feedback, not class A operation.  I'll bet that Erb amp wud clean up really well with a good 10db wrapped around it. There's plenty of gain to work with, esp with those little chip pre-pre-rivers. Let me know if you know how to do it, Frank.  Maybe it's as easy as adding a single toroidal transformer from the output back to the predriver, out of phase.

T

[WA1GFZ]

If you have 50 ohms at penny and 50 ohms at the pad then the  penny is looking into 25 ohms. Take one of them out.
Each Erb stage has negative feedback. Too much phase shift over the frequency range to wrap it around all three stages. I've measured well above -32 dB IMD3 but never tried it with higher bias yet.
What is the current in the first stage at 16 volts you might be heading into class B land.
You will not improve imd until you bias the stages for low IMD and fix your thermal problem. Those FETs are very clean when biased properly, I would assume they needed to be clean for MRI service. The digital bias control keyed the bias on just before a pulse was generated to save power/heat.
I think Jay will agree with me that erb is designed very well.

[K1JJ]

I've brought the VCC up to 33 volts several times testing to be sure everything was idling hard. Set the bias pots until the driver and finals were warm to the touch.  Still, even with a smaller 10w out, it was not as clean as the FT-1000D.  I saw the same wid the CCI amp.  Sure it's acceptable for ham use, but I'm trying to go a step beyond that.

I've had poor luck with SS amps for cleanliness so gonna give the tube amps a go for now.  I have an old TMC 100mW in, 6CL6 > 6146 > 4CX-300 amp with neg feedback that says -40db 3rd in the manual. I might stick in a pair of 6146's in place of the 4CX's and have no blower noise. EZ 30w pep out, Add some more NFB and see how it plays.  No LP filters needed.

Either that or the FT-102, not sure yet.  I'm done with the SS amps for a week anyway.

T

[WA1GFZ]

Well, your choice is high Q tuned circuits or high conduction angle.

[K1JJ]

Back to testing the solid state Erb amp.

Had it on the bench running. After an hour of success, the thing suddenly flashed and took out the two finals and drivers while adjusting the bias. I put caps across the pots and will not adj bias on the fly anymore. The pre drivers are OK. Fragile stuff. 

I'm replacing them using FETs from the other Erb and CCI amp.  Got a fan going on it and a 24V supply for the predriver and a 40V supply for the driver/finals.  The MRF-148 appears well rated for 40V, right?  I see 125V rating IIRC.

I see the specs for MRF-150's , etc show -50db 3rd IMD, so worth another try.

T

[WA1GFZ]

Tom I think the high voltage (40) is only on the final but don't have the schematic in front of me. Anyway sounds like it took off on you.
I was looking at the erb and I think we can add more feedback around each stage pretty easily. Need to study it some more. Maybe we should add some temperature compensation on the bias. Just bolt a diode to the heat sink and put it in series with the bias pot on the ground end leg. As the diode heats up the junction voltage drops reducing the bias. Turn the pot down because it will offse the bias about .7V higher.

[K1JJ]

Gots it working again.

I can get out 150w when pushed at 45V VCC.  But I backed it off to 40V for tests.

When I idle the MRF-150's in maximum heat class A, the waveform looks FB up to 50w. But as I drive it into AB2, the waveform shows signs of compression/clipping on the peaks. It's the finals cuz I checked the driver and predriver and they look OK.

Does yours show this slight flattening of the peaks when it AB2?  It looks like saturation, though the amp will put out another 100w after the first clean class A 50 watts.  If I idle it harder, then I can get more clean watts out. For some reason it doesn't sem to like gate current in the finals.  The 24V and 40V power supplies are totally stiff with no drop at all from zero signal to full signal, even at 150w out during this compression effect.

I'll run some more tests and see how the IMD at 50w class A loooks now at 45V.

T

[K1JJ]

Nope, the Erb amp just ain't cutting it.  I drove it with the clean SDR. 

It's slightly better at higher voltage, but not much. I brought VCC up to 50V with a hot idle on the MRF-150's. The drivers are running very warm too. I can almost fry spit on them after 20 seconds.  Even at only 20W pep out, in full class A - stable drain current,  it's still is no better than -30db 3rd.  When I hit it harder, like for 75w AB2, it degrades down into the mid to low 20's.   The predriver is at 24V and the driver/finals at 40-50V. The driver at 24V doesn't change anything.

There's different Fets in there now, so I doubt there's a problem with them.

It's putting out the soup and is stable with no parasitics or trash. Just that it is rough around the edges.

To see another 20db 3rd IMD improvement would take something drastic to be changed I think.  No hotter idle will improve this as-is.

As a comparison, the barefoot SDR driver is so clean I can barely hear side products at all - like they're not there.  All measurements are normalized first at +40db over to prevent receiver overload..


Any other ideas to try?

T

[WA1GFZ]

Tom,
I was working on the transient and drop out problem tonight. Go into your set up  DSP tab and reduce the buffer size to 1024 for both RX and TX. Then go to the audio Tab Primary and reduce that buffer to 256. This seems to fix things and allows me to run at 192K sample rate. Horsepower went up to about 16% from 7%. I need to do more testing but this seems to work.
I wonder if the erb needs a low pass filter. I'm looking at adding more feedback but have not decided which way to go. I need to build up another power supply before I can get into that. I do remember looking at the gates of the finals and it seemed distorted compared to the driver. Maybe the transformer cores are saturating. I was thinking of rewinding them with two cores stacked at one point but Jay thought they were fine. I do agree with Jay that type 61 material was fine and I did not see much change when I swapped over to type 43. It looks like there is enough room for extra cores to lay flat on the board. Type 43 saturates easier than type 61 so that could be the reason it didn't change anything. I would rewind the transformers with twisted teflon wire

[WA1GFZ]

We could just use cores off dead erb boards. Do not use any predriver drain cores. They are type 43 material. We need to be careful with the maximum gate voltage rating on the MRF150s. We don't want to drive them into failure.
Do not try to pull more than 100 watts output at high bias.
Try 40 or 20 meters to see if the gate voltage looks better. This will tell us if the transformers are saturating on 75M. Maybe the inductance is a bit too low and stacking cores will fix that.