EMRFD Message Archive 11203

Message Date From Subject
11203 2015-06-16 21:13:48 Ashhar Farhan A low distortion QRP linear
my search for low distortion linear has culminated in a very easy 3.5 watts linear amp.

EMRFD sets the context for any such attempt. I quote (sec 6.6, second page)

"The design begins with a pair of equal IF signals or two tones"

Hence, using a two-tone generator, I started to measure the IMDR of various devices in an FBA varying the emitter resistor to control the device current. The 2N3904 turned out to be a non-starter beyond +40dbm.  Then, I switched over to 2N2222A. I was pleasantly surprised that it could be pushed to +45dbm. 

However, curiously, paralleling them didn't seem to improve the IMDR. I paralleled up to eight devices, each drawing 60ma of current. The IMDR stayed solidly at +45dbm. can someone double-check on paralleling devices to increase the IMDR? I have verified these results. I was expecting at least 3db improvment in the IMDR when the devices were paralleled. The improvement was marginal beyond the second device.

Using this data, I played with cascade.exe. For those who haven't tried this gem, I highly recommend it. It allows you to plug in the noise figure, intercept and gain of a a number of cascading stages and measure the result. I can't recommend it enough.

It turned out that the stage gains were crucial. often, more gain early on resulted in distortion from the previous stages that propagated to the following stages. hence, the thumb rules to have emerged are :

1. Keep the number of active devices in the chain less.
2. Keep the gain low towards the input and more towards the output.
3. Pads are bad for IMDR, use them judiciously.
4. It pays to have high IMDR stages even early on. Go ahead and pour  50 ma into the drivers as well.

Finally, I used an RD16HHF1 baised for 700ma. Ouch, I know that that is a lot of current. But take a look at the scans at https://groups.yahoo.com/neo/groups/emrfd/photos/albums/646547507

IMDR is better than -40 dbc
harmonics are down by almost -30 dbc without any harmonic filtering! I suspect that the harmonics willl improve even more as the twon-tone generator has the second harmonic that is barely -40 dbc .

This one is eminently suited to being driven directly from a filter that follows a diode mixer. It needs just -18dbm input to produce 3.5 watts output. Keep the attenuator pad at 6db.

Surely, if u shout the power can jump to more than 7 watts. But  go easy, people will hear your clean signal and bless you.

The amp hasn't specified any harmonic filters as I use a common set of harmonic filters now for all my rigs. 

- f

11204 2015-06-16 21:18:17 Ashhar Farhan Re: A low distortion QRP linear
I forgot to add, for the final transformer, the TV baluns didn't work. I must repeat my measurements with parallel devices with FT37-43. At the level of a few watts, the ferrite's IMDR starts coming into play.

- f

11205 2015-06-18 11:50:56 w0gk Re: A low distortion QRP linear
  Interesting work!  This is an area that I'm interested in, and have done some experimenting. It seems like there are a lot of published designs for both lower and higher power linear amplifiers, but not a lot of designs for getting from 10mW or so up to 2 or 3 watts of very clean power.

    Chris Trask, N7ZWY has a paper at
http://www.home.earthlink.net/~christrask/Bipolar%20Transistor%20Evaluation.pdf
in which he evaluates various bipolar transistors for linearity. It might give you some ideas for other devices to try.

73, Randy, W0GK
11206 2015-06-18 12:24:10 Ashhar Farhan Re: A low distortion QRP linear
ron pointed out that i have skipped drawing the bias circuit for the final FET. it is just a 10k bias resistor from a 10k trimpot across a 78L05. standard bypass rules apply.
thanks for the heads up about trask's work. reading now...
- f

11207 2015-06-18 12:49:23 w7zoi Re: A low distortion QRP linear
Hi Farhan and gang,

Great work.    This is just the kind of thing that I like to see.   That is, it's measurement based along with analytical work.    The combination can be extremely powerful.

It's also great to see you moving toward spectral purity.    That's something that we tend to forget with those first home-brew SSB efforts.    A local buddy (WA7MLH) recently built a rig with a 10 element crystal filter with a bit wider bandwidth than usual plus a carefully done low distortion amplifier chain.   He commented that the rig produces all sorts of surprises from the folks he works with it.  They just can't come to grips with the fact that a home-brew rig could sound so good. 

You mentioned coming up against an intercept wall with some parts, or with parallel combinations.   Another vital part to the design is the output match of each stage.    You were running a driver with a 2:1 turns ratio transformer in the output, so the collector load is 200 Ohms, assuming something has been done at the input to the RD16 part to generate 50 Ohms.    But 200 Ohms at the collector of the driver may be too high.  Dropping back to 50 Ohms might give you the needed OIP3.    It's instructive to look at the collector waveform with a scope while doing the two tone testing.   If the transistor gets close to saturation, you can be sure that there will be problems.    A design from a very long time ago with a pair of 2N3553s in parallel (EMRFD Fig 6.98) produced a measured OIP3 of +43.5.    This is not all that good, for the amplifier was intended as a 1 W output circuit.    It was, however, pretty good if not driven to full output.

Keep up the good work.

73, Wes
w7zoi

11208 2015-06-18 17:41:36 kb1gmx Re: A low distortion QRP linear
Good so far.

Suggestions, many.  Try a push pull output with decent standing current
as its more efficient and can be cleaner.  A pair of RD06s runninng 150-300ma
each as a start. 

 Use transmission line style transformers to keep flux loading of the ferrite down.

Make the driver push pull with something in the 2n2219 class and use 
100 ohms or less for collector loading.

Even with a class A driver 200 ohms is too high for any power and invites votlage
clipping and nigher non-linearity.  It will improve bandwidth as well.

The gates of power fets are problematic as they are not an impedance match 
and also subject to Miller effect through Drain to Gate feedback.  

The use of negative feedback is worth considering as power fets have high gains
and it can help with linearity.
 
In all cases you want to keep the collectors/Drains from seeing large excursions in
voltage.  This means devices that work well at higher currents.  Or parallel devices.

Try to avoid emitter/source degeneration on higher power deivces it can lead to 
problem in driving the device.

Bipolar devices require stiff bias sources, the bias source should behave as a true
voltage source at all levels of drive.  If the RF drive can modulate the bias current
then linearity suffers as the transistor is departing from the desired operating point.

Same applies to fets.  The voltage source should be stiff and delivers through as
low a value resistor to insure the bias is stable.  That resistor can also serve
as gate loading.

The DC rails should also be AC stiff.  This means excellent bypassing and 
if need be regulated.  If one stage can under power modulate the DC supply 
the drivers and earlier stages may also suffer IMD as a result of changes 
in operating point..


Allison
11209 2015-06-19 10:48:19 Ashhar Farhan Re: A low distortion QRP linear
Allison, Wes,

Thanks for the great suggestions. I suppose I must try out lowering the impedance of the driver ... The linear is already in-situ in the Minima. No contacts yet as my balanced line tuner is under works at the moment.

I must also experiment with the MOSFET feedback. It will be interesting to see the impedances around the final with the feedback...

I must say here that I discovered the approach to making low distortion linear in just under a page of 6.54, under the topic 'Linear Power Amplifier Chains'. It places the approach very simply as 'Design begins with a pair of equal IF signals, or two tones'. That's almost all I needed to know. 

The other tool, that finds very little mention about us (and hence I extrapolate very little usage too) is the cascade.exe on the EMRFD cd. this allows you to figure out how linear amp chains, receiver front-end et al will perform. You plug in the noise figure, intercept point and loss of each stage and it does all the calculations for you. You can change any of these for any stage and watch how the effect ripples through. It is a great tool.  I have used it to design the new minima receiver and it is really an excellent sounding receiver with a very low component count. I spent hours experimenting with the receiver line-up on cascade.exe. it almost addictive to try out these castles-in-the-air.

I would like to return to the bench this weekend and pursue the IMD performance of paralleled devices. I assume that each device, would contribute half the actual output power and hence, operate 3db lower, taking the IMD down by 9db. 

I tried paralleling upto 10 2N2222As in class A for the final (in place of the RD16HHF01). Each transistor was drawing 60 ma. The idea was to pour in 600 mA in the hopes of getting a decent 2-3 watts output with lower than 40dbc IMDR. The IMDR remained at -20dbc. But now, I suspect the poor quality of TV balun as the output choke being the culprit. The BD139 was


11210 2015-06-19 11:36:38 kb1gmx Re: A low distortion QRP linear
60ma per device is too high for 2n2222 series.  The problem is you need a large area device to absorb the current and not suffer internal degeneration to the resistance [especially the emitter] in the junctions and leads.  That and many devices at high 
current suffer saturation effects [they run out of minority carriers] that lower HFE.

FETs are majority carrier devices but there are other headaches there with high capacitances due to large area junctions [or in the case of some many paralleled 
devices internally].

In all cases with increasing current you also go to decreasing impedance in the output circuits though often with improved stability at the same time.

As you experiment with cascade.exe you may also find that once you get to a level the quality of the source signals are under consideration.  The reminder there is GIGO
where garbage in results in garbage out.  

Also the noise figure of the TX amp chain starts to take on a life of its own. Its pretty 
scary when the cumulative noise (phase and thermal) with enough gain may be 
-56dbc but at 1kw that's a lot of signal going to the antenna.  Turns out that many 
large area bipolars are fairly quiet.  Its the low level amps that create and amplify 
the noise.

Its an interesting area to explore as some have observed, we are getting phenomenal
receivers that suffer from less that stellar transmitters, often in the same box.

Allison

15520 2019-10-15 13:47:33 prysm678 Re: A low distortion QRP linear
the photos regarding this experimental are no longer accessible
is it possible to repost?

regards
fernan, ve4feb