RF Feedback

[AB2EZ]

Bob

et al.

Based on the analysis I did in my previous post... I can't see how R4= 82 Ohms could possibly deliver the required average base current when the amplifier is putting out significant RF power. This, by itself, would likely produce a feedback mechanism between the output and the input of the transistors... leading to instability.

Stu

[W1RKW]

I kind of think the same Stu based on some empirical poking yesterday.  I dropped the R4 value to about 70ohms to turn on the  transistors a little more.  With R4 being 82 ohms they were floating around .64Vdc.  Bumping them up closer to .68 seemed to turn them on better but also changed the oscillation rather than eliminate it..

[AB2EZ]

Bob

I think that this version (with R4=82 Ohms) is not appropriate for AM.

The 470uF capacitor across the bias source (the diode) can act as a charge resevoir for brief periods of time during modulation peaks in SSB mode... allowing the base of each transistor to draw more current than R4 can deliver, for those brief periods of time... and the charge that is removed from the capacitor will be restored (refilled) via the 82 Ohm resistor (R4) during periods of relatively low RF output.

However, the 82 Ohm resistor cannot deliver enough continuous base current to keep the amplfier at carrier in AM mode.

I think you need to revert back to 33 Ohms... and after making this change, you need to check that the resting collector current (no RF input) is not too high.

Stu

[WA1GFZ]

Bob,
Try .01 in series with 100 ohms (2W) between base and drain of each transistor.
A little NFB should tame it. I've found diode thermal bias compensation not the greatest. It will still run away if the devices get hot so best to starve then a bit and let heat bring up the current. You can increase the collector to ground caps if you don't plan to use it on 10 meters. Make sure the Bias diode has a good thermal bond to the devices
I played with the MRF429 enough to prefer the mRF150 FET. The PCB layout and NFB are very critical using bipolars if you want a BB flat input Z.

[WD5JKO]

I just looked at my modified 10 meter amp after conversion from 11m done sometime the last solar cycle. I made some changes I don't remember, but that is another thread, under CRS.  I did the NFB just as Francis had suggested (and Shane) to lower the gain, improve stability, and lower IMD. I also put a 510 ohm 2W resistor across the RF amp output so during standby periods the amp at least had some load instead of an open circuit. The EB63 also leaves the AMP unloaded when idle, and biased up too!! 

I see considerable rework with the bias supply done (CRS though), and I ended up with the bias source shorted out (no bias). The amp as I recall worked just fine on AM (3w in 20w out AM).

After reading Stu's analysis of how the EB63 stock bias is unsuitable for AM, I understand why I eliminated the bias on my amp. For AM, NO bias is preferable than a high-Z bias source. With the EB63, changing that bias resistor from 82 to 33 ohms is likely to raise the idle current too high, increasing the 12v draw, heat, and lowering efficiency.

Since we seem to want to preserve the bias with this amp, we need a low - Z bias source. An idea comes to mind and I will outline the concept here. Use a LM-317 adjustable regulator which provides 1.25v when the adjustment pin is grounded. Then in series with the 1.25v low-Z source put a series resistor, maybe 10 ohms to the bias diode D2. Now the threshold is gone, and the Rs of the bias supply is 10 ohms instead of 33, or 82 ohms.

We can vary the idle current as before by playing with the series resistor value, or maybe make the LM317 adjustable over the range of 1.25v to 2.0v.

I was just typing as I was thinking, so hopefully this lower Z bias source concept will work.

BTW the term CRS means, "can't remember sh_t" 

Jim
WD5JKO

[AB2EZ]

Jim

I'm beginning to think that your suggestion to just ground the center tap of the secondary of the input transformer (no fixed bias) is the simplest solution for AM operation, if the modulation index (for negative peaks) is not too much greater than 50%.

The more I think about, the harder is seems to be to bias the transistors, properly, to operate in Class AB or Class B with an AM input signal.  

With (for example) a 2 Watt input (at carrier), you have 14 Volts peak (assuming the input impedance is close to 50 Ohms). The 4:1 (turns ratio) input transformer will reduce this to 3.5 Volts peak (i.e. 1.75 Volts peak from each side of the secondary to the center tap/RF ground). This is probably enough signal to drive the transistors... while still providing reasonable linearity (RF amplitude out v. RF amplitude in) on up to 50% negative modulation peaks. I.e. (1- 0.5) x 1.75 Volts = .875 Volts

What you really want to do (but which is hard to do) is:

 A) to inject a fixed amout of current into the base of each transistor... adjusted to produce the desired resting collector current in each transistor. I.e. you really don't want a fixed voltage... and trying to use a fixed voltage will lead to problems such as the thermal runaway problem that Frank mentioned.

and

B) to provide a low impedance DC path from center tap to ground... in order to provide the additional average base current, as needed, when there is R.F. applied to the input.

As a compromise, one might redesign the biasing circuit to employ a 33 Ohm (or lower value) resistor... but to use a biasing voltage that is closer to 0.5 Volts (A Schottky diode that can handle the current?). That way, the transistors will be operating closer to Class B (but still a little bit into Class BC).

Stu

[WA1GFZ]

Changing the bias will also effect the input Z of the amplifier. I would put a shunt resistor load in parallel with the thermal diode so the voltage can't raise high enough to hurt the devices. I would set the bais so there is just a bit of collector current close to class B. This way as the parts heat up the diode will draw additional current off the bias. When diode cools down the resistor sets the maximum voltage. I found this worked best when playing with MRF429s

[W1RKW]

Frank,
I have no intention of running this amp higher than 75 meters. So if narrow banding it is necessary, I won't lose any sleep over it.

I blew a transistor by messing around with what I thought was NFB.  Not sure that I was doing it right.  If anything I only improved the oscillation to the point of destruction.

As soon as I get replacement transistors I'll tinker using some of the above suggestions.

[WA1GFZ]

A current limied power supply is your friend

[W1RKW]

OK, got new transistors today.  Installed them and started troubleshooting again. 

Put a 0.01uf from R4 (82ohms) to ground.  It did nothing to suppress the oscillation. 

Tried 2  0.005uf from base to emitter on each transistor.  They reduced the amplitude of the oscillation by 10%.

In standby mode when oscillation isn't under load, putting a .005uf across the primary of the input transformer kills the oscillation.  The same with a jumper. 

I quit for now.  Will come back to it tomorrow for a brief stint. 

Need to try to get some rest. Been running full tilt the last 3 days with minimal sleep and don't want to screw up.


 

[WD5JKO]

Yes best to go slow, and deliberate when you are not tired.

So in my opinion, stabilizing the AMP when the relays disconnect the inputs and outputs should be attacked first. Here is what I'd do:

1.) Remove the class AB bias (short out D2) during receive. Could add a relay, or figure a way to use a 2N7000 fet. No since having the transistors sucking current while you are receiving. In fact they might create noise causing poor reception problem.

This should quench it, but if not, then:

2.) Add a 510 ohm resistor (2W) across the RF output (of the AMP) to always have some load on the AMP.

If this is still less than 100%, then:

3.) Add a 510 ohm resistor across the input (of the Amp) to always have an input load on the AMP. Better yet, install a 3 db pad at the input since the amp has too much gain anyway.


Then when you finally get to running the AMP from the retro, to get the gain & linearity right you will likely need NFB to achieve both variables. Shane suggested adding NFB early in this thread. I just looked at my old 10m amplifier, and I did the same. see attached picture. Notice the Amp already had 47 ohm base-emitter resistors. The R-C I added from C-B on each transistor used a 2200pf - 47 ohm 2W. For your AMP on 80m. the cap should be around .01uf, and I might start with ~ 220 ohms, and work my way down till the gain and linearity are right to match the Retro-75.

The AMP pictured would do 20W carrier AM, and ~ 80W PEP. The bias was OFF all the time since it caused more problems than it benefited. I don't recall all the details, but I did a lot of modification in that area, and ended up jumpering out the bias. That meant NO SSB however.

Jim
WD5JKO

[AB2EZ]

Rob

Putting 0.005uF across the primary of the transformer is equivalent to putting .005uF x 16 from base-to-base (but not base-to-ground). 0.080 uF has an impedance of around 2 ohms at 1 MHz. This is forcing the inputs of the two transistors to be closer to being in phase... and since their outputs are connected out-of-phase, it is killing the oscillation.

The question remains: what is the cause of the feedback.

I think you should, next, try reducing R4, the 82 Ohm resistor, to something like 33 Ohms... and also add a 0.1 uF (or larger) r.f. bypass between R4 and ground (in parallel with the existing 470uF charge reservoir capacitor).

Stu

[AB2EZ]

Bob

One other thing...

It is possible that feedback from the output is getting back to the input via the primary of the input winding (and the various things that are connected to the primary of the input winding... like the automatic antenna switch).

Try flipping/transposing the primary leads.

Separately... even if you stop the oscillaton by eliminating or cancelling the feedback... you still have to change R4 to 33 Ohms if you want this amplifier to work on AM (rather than just SSB).

Stu

[W1RKW]

any idea what the B-E current of the MRF454 is. The data sheet doesn't provide that.  I'm leary of changing the 82 ohm resistor to 33ohms .

[WA1GFZ]

Do not increase bias current unless you want to buy more transistors.

[AB2EZ]

Bob

You can check the current gain of the transistor. The Base-to-Emitter current is (1/the current gain) x the average collector current (at carrier). According to the MRF 454 specification sheet, the DC current gain (H_fe) is between 40 (minimum) and 150 (maximum).

If you use a value of 100, for example, the average base-to-emitter current has to be the average collector current /100.

Let's assume that the average collector current for each transistor is (for example) 3 amps at carrier. I.e. 3 amps x 12 volts =  36 watts per transistor of electrical input power, at carrier... corresponding to around 12 watts of r.f. output power, at carrier, in Class AB linear operation (per transistor).

That means that the average base current has to be 3A/100 = 30mA per transistor => 60mA total average base current.

Add, to that, the average current drawn by the two 10 ohm resistors (one from each base to ground) = 2 x 0.7 volts /10 Ohms = 140 mA

So the total current is 200 mA (not including the current through the biasing diode)

200mA x 82 Ohms = 16.5 Volts.. i.e., you can't deliver the average base current to the transistors and the two 10 ohm resistors with an 82 ohm resistor fed by a 13.6 volt supply.

You can do the job in sideband mode, because you don't have to deliver the constant average base current associated with the AM carrier.

This is why they used a 33 Ohm resistor in the original design. The revised design uses an 82 Ohm resistor... but it is specified only for use on SSB (not AM).

Try a gradual reduction of that resistor's value... e.g. put 150 ohms in parallel with it to get a combined resistance of 53 Ohms


Stu

[W1RKW]

I 1/2'ed the 82ohms to 41ohms. I didn't have a 33 on hand.  Brought the DC up slowly and got to about 11.0V and heard a phfft sound and immediately killed power.  Not sure what it was  but feared the worst.  So I pulled the transistors and checked them including the bias transistor.  All seem OK for the most part but I do question the MRFs.  Even though the BE and BC junctions check good CE readings are all over the place and  in most cases around 50K but waver. Sometimes they're as low as 37K and as high as infinite. I don't get a consistent reading on either transistor and neither behaves the same.  Different meters give different results.  Tried my old VTVM and it shows both CEs being shorted.  I don't know what to think.  If I turn off auto range on the DVMs the CEs show hundreds of megs so I'm inclined to think the transistors are still OK and the phffft sound was the relay possibly beginning to chatter because of oscillation and the COR. Though the COR has never reacted to the oscillation.  Clues?

[W1RKW]

revisit.

May have a clue or 2.

tried it again but took it in steps. 

At 13.6V in there's about 155mA through the 82 ohm resistor.

cutting in half the value of the resistor to 41ohm with 155mA going through that value there should be about 6.3V across the 41ohms.  So advancing the PS so that the 41ohms is approaching a 6.0V drop the relay attempts to close.  Oscillation is very strong at this point and is activating the COR.

At 82 ohms the oscillation does not occur until the transistors are on.  And that is what is occuring with 41 ohms.

The next step is to try flipping the transformer and see what happens.

[WD5JKO]

Bob,

  Since the AMP input and output is unloaded with the COR relay not energized, I think having those transistors all biased up in class AB mode like this is asking for trouble. Sure if you really want to do this, then only apply the bias when the COR is energized. This way both the input and output sees a load, hopefully in the order of 50 ohms.

  Maybe my way of thinking is flawed, but with the I/O unloaded (COR not on), isn't this similar to putting your car in neutral, and then putting a brick on the gas pedal?

BTW I just received my Retro75 + case kit. OH Boy!!

Jim
WD5JKO

[W1RKW]

Hi Jim,
I tried no bias on the transistors.   I ended up with a squarewave generator when RF was applied. No bias certainly squelches the oscillation but when driven there's clear indication of the oscillation on the output.  I need to build an attenuator.  It's possible I'm driving the piss out of the amp with the Retro.  The oscillation does show up when RF is applied and COR is activated by RF but does not appear when the COR is operated manually with no RF applied.  A 50 ohm load on the output does kill the oscillation but is evident when RF is applied.  Not sure what is occurring at this point.  There's never enough time to tinker.

[WD5JKO]

How about NO bias only when the COR is not energized? That would mean interrupting the bias resistor 82 ohm, or 55, or whatever with an additional contact on the COR circuit, or some solid state means to accomplish the same thing.

Letting a high gain RF amp IDLE in AB mode when the I/O SWR is infinity (open circuit) is not something usually recommended. The other thought is to put two 510 ohm resistors, one across the input and and the other across the output (size for the power levels). Now the amp sees a 10:1 SWR instead of infinity and beyond.. 

Just a thought...

Jim
WD5JKO

[W1RKW]

yep, tried the no bias no RF drive and manually activated the relay.  No osciallation.

Also, no bias and no RF drive (stand by), no oscillation as expected since tranny's are cutoff.

It's when there's drive and bias that it oscillates. And of course it oscillates when there's bias and no RF drive as well.  I'll have to try the 510ohm trick.  Stand by.

So have you started your Retro yet?

[WD5JKO]

So have you started your Retro yet?

Nope, just came in UPS. Getting the enclosure also gets you all the pots, connectors, etc. It was professionally silk-screened too.

Getting ready for a two week business trip to the Beverly, Mass. area. I could take the kit and do it at the factory I work at over the weekend, or instead go to Gloucester and take a "whale watch" three hour tour. I did that before, on a choppy day. After most of the passengers lost their lunch, and turned every shade of green, I ordered a greasy hot dog with all the fixings. Washed it down with a beer. A few folks after seeing me eat the dog had to go back out and repeat...

Jim
WD5JKO

[KC2YOI]

Sounds like the latest Hollywood diet.... FFFffwwarff 

[Gito]

Hi

From the ARRL hand book
 LOW-Frequency Parasitic Oscillation ...bi-polar transistor exhibit a rising gain characteristic as  the operating frequency is lowered.
To preclude low frequency instabilities because of the high gain,
Shunt and degenerative feed-back are often used.Feed-Back (from collector to base) increases by nature of the feedback net work reducing the amplifier gain.( depends on the Frequency}
the feed back net work consist of R in series with C from collector to base of the transistor.
The value of C and R are usually chosen experimentally.C is usually between 220pf and 0.0015uf for HF band amplifiers,the net work has small effect at the desired frequency but has a pronounced effect on the lower frequency.
the R is between 52 to 5600 Ohm

Using R between emitter and ground develops a degeneration at low frequency ,and bypass it with a C for adequate RF bypassing for the intended operating frequency.The impedance of C rises progressively as the frequency is lowered   .
this R is seldom greater then 10 ohm and maybe as low as 1 ohm.

Using swamping resistor in the input amplifier usually from 3 to 27 ohm ,placing it as closed at the terminal base.


If You have the ARRL book 1997 edition, ,You can find this article in page13-27 and 13-28 more compelletly


Gito