Wellgood balanced loop amplifier, (Wellbrook clone).
edited February 2018 in Installation, Antennas and Interference
I have the first Wellgood LNA online @ ka7u.no-ip.org:8074 . It is using a 120cm loop made out of 1" (25mm) soft drawn copper pipe. One of the transistors is running hot and might not last so may not be done with it. Hi Hi. Earlier this evening I was copying Jim Creek down around 24.8KHz quite well, but it is not loud now. Could be the solar storm or it could be the hot transistor. I was also hearing Radio Thailand well at the same time @ 17.630MHz so the frequency range seems to be there. I'll evaluate it to the W6LVP and Hiletgo loops after I figure out what is up with the hot transistor or after I get another board finished. Winding and soldering in the transformers is a serious chore. I substituted parts from George Smart's list to fit the parts in stock at mouser.com and chose to use 2N3866 transistors in the metal can. I'll leave it on the KiwiSDR for a while unless it burns out.
This is funny, It occurred to me that this amp with these transistors might not want 9-12vdc to run. So adjusted the voltage for best noise floor with good signal peaks and find that 3vdc is close to optimum. It actually works well at 1.5vdc. How about that? And the transistor seems to be very happy too. I think I'll need to readjust the balancing resistor, although I doubt lowering the voltage will affect that setting.
I changed transistors on the 3 Well good boards and now they keep a low noise floor over a range of voltages. With these transistors the gain starts dropping at about 10 volts. The optimum voltage seems to be 5 volts.
Unit did not work on initial power up.
Comparing Dr. Smarts schematic to his Toroid winding drawings revealed a glaring
Error - note attached photos.
George's drawings don't jibe with the schematic - according to his drawing he has the RED and BLUE single turns on Toroid 1 being soldered in the wrong place - and the error ties the two emitters together - not harming them but rendering the circuit inoperable.
So: Single TURNS "Red" and "Blue" should go tie points (in example) BLUE PAIR wires on 1 and 2 And RED PAIR wires on 6 and 7
or VICE VERSA.
Of course if I was on the ball, I would have spotted the error and wired the toroid correctly -
Now I can remove the entire toroid, unsoldering everything OR cutting the traces under the board and cross wire the lines the way they are supposed to go. Thankfully I have a HAKKO rework station with a good de-solder tool.
and, hey, if I am off on this please let me know.
Colin - VA7WWV - DXer.ca
(owner of 3 Wellbrook ALA100's and 2 FLG100's - maker of Dual Balanced antenna phasers and VACTROLs for FLAGS
I think George's drawings are correct.
Some time ago I asked him about this apparent mistake. It turns out that although he has crossed the secondary windings over, he has also changed the feedback winding nomenclature to illustrate this. He did this to make it easier to pins 3 & 5 as you see in the instructions for winding T1 on the page. It's less fiddly to implement than crossing the feedback windings. It makes winding the transformer easier, and essentially only flips the DC ground connection to the two transistors, which will not matter.
I agree it's not immediately obvious and it certainly caught me out too.
Personally I prefer the LZ1AQ design as it's easier to build and I think it's also a better performer.
I didn't bother with the voltage regulator or input low pass filter C5 / C10 / L1 /L2 and changed the output transformer winding ratio so that I could feed it with coax.
More info here.
Martin - G8JNJ
That and maybe drawing out the actual circuit diagram because the schematic on Dr. Smart's webpage is not quite correct - and led to part of the confusion -- but not as confusing as my mistake! :-)
Thank you for the quick response - Dr. Smart was very helpful and good humoured.
Good luck when trying to measure the IMD performance, especially using the single tone method.
Here are my measurement results, note that I'm using amateur grade kit to measure it, so there may well be errors creeping in, but I have spent more than 6 months gradually optimising the test setup, so that I can obtain repeatable results and at the very least 'compare like with like'.
Note that the calculated values using the two tone and single tone methods don't match, and the difference varies depending upon the amplifier configuration, so its not possible to apply a simple formula to bring them into agreement.
Wellbrook quote an OIP2 of +70dBm and an IOP3 of +40dBm for their 'old style' amplifier which the Wellgood is based upon, and an OIP2 of +90dBm and an IOP3 of +55dBm for their 'new style' amplifier.
My results indicate the Wellgood to be somewhere around an OIP2 of +57dBm and an IOP2 of +37dBm, which is not a million miles away but could be better.
It maybe that if I'd used slightly bigger ferrite cores and optimised the bias I could have matched Wellbrook's figures, or it could be that George has not quite got the transformer winding correct WRT the way the overwinds are placed on the binocular core with its three (possible) separate magnetic paths.
Bottom line the LZ1AQ is better than the old style Wellbrook, but the new style Wellbrook may be as good as, or better than, the LZ1AQ.
Martin - G8JNJ
Great for aggressive testing in hostile outdoor environments, cheaper to build and if something breaks I can fix it, unlike the Wellbrooks which are potted in resin and basically toss in the bin if anything gives out.
My Wellgood amp is constructed on two small pieces of veroboard supporting the major components attached to a copper pcb sheet as an earth-plane.
I used BC550C transistors, similar to 2N2222, and it works well and is stable , but I have had to modify the base bias divider resistors because the current is rather high with these devices using the component values given and with 12 volt supply.
A quick calculation of the design supports this:-
Bias divider R2/R1 2K ohm and 1.2K ohm, therefore voltage at junction = 1.2/(1.2+2) x 12 = 4.5v. on base of transistors.
Voltage at transistor emitters approx 4.5v-0.7v = 3.8v.
Current through each transistor = 3.8/50 x 1000 mA = 76mA., total current drawn just over 150mA.
Dissipation needed for each transistor, assuming perfect balance = V x I = (12-3.8) x 0.076 = 0.62 watts.
If my rough calculations are correct, using 12v supply this is too much dissipation for small transistors like the BC550/549/548/2N3904 (and plastic 2N2222 ?) which are rated around 500mW with limited heat conduction from their junction and therefore get unacceptably hot with 76mA, particularly in a confined space without a perfect heatsink. Perhaps a 6v supply was used originally?
Also, at 76mA current, the 100 ohm bias potentiometer is dissipating 580mW, so is a little under rated.
Using the design values and 12v, a metal cased 2N5109 or 2N3866 or similar would seem to be more suitable with 2W dissipation rating, but they are becoming more expensive in their discreet form. However they certainly have less noise and higher overload margin with the higher current, but it would be nice if a small cheap transistor like the 2N2222 or BC548/549/550 could be used without affecting the practical results adversely under normal receiving conditions.
For my BC550s I decreased the amplifier current to within their rating by changing the bias divider resistors to 4.7K ohm and 1K ohm , leaving the 100 ohm potentiometer for emitter bias as it was. The calculation for these values is:-
Voltage at junction R2/R1 = 1.0/(1.0+4.7) x 12 = 2.11v..
Voltage at transistor emitters approx 2.11v-0.7v = 1.41v.
Current through each transistor = 1.41/50 x 1000 mA = 28.1mA.
Dissipation needed for each transistor, assuming perfect balance = V x I = (12-1.41) x 0.028 = 0.3 watts. (well within the rating of 500wW)
I am now measuring a current of 62mA with 12.3v supply so the calculation is about correct, adding a few mA for the voltage divider.
The transistors still get very hot to the touch, but with a heat sink they are just about tolerably hot (40C).
Perhaps it would have been better to increase the value of emitter bias resistor? This could be by inserting a fixed resistor in each emitter circuit (bypassed with capacitor). This could be adjusted to suit the dissipation rating of the transistors used.
Flat topping starts to occur at about -2dBm, which is "fair". It improves slightly as the current is increased but not too much difference.
The noise factor using a calibrated noise source at 50 ohms input to the amplifier measures 7.5dB , but may be better with the low output impedance of the loop.
Regret I have not the test equipment to do OIP tests.
Measuring the gain and response with a VNA, the amplifier has almost 24dB of gain at LF, up to 20MHz, dropping off to 14dB at 28MHz. Nice high gain for a relatively wideband amplifier at least up to 20MHz when it starts to drop down. (Twice as much gain as a conventiional Norton single transistor amplifier which has about 11dB gain with typical transformer ratios).
On reflection this may not be a bad response, if the 1m loop is relatively more efficient at the higher frequencies and may need less amplifier gain?
Marginally more gain at 30MHz if the 82pF capacitors are removed but I left them in place for when I try the loop.
When winding T1 I used a thin PTFE sleeve in the core for the secondary. The primary and feedback turns were directly through the core.
To make the output transformer T2 much easier to make, I used a larger BN73-202 binocular ferrite with half the number of turns (i.e. 4+4t : 3t). Two stacked 303s would probably be similar. The frequency response still extends down to 100kHz with 22dB gain. I also used 1uF for C3, and will use 220n or 470n for C7 when I make the loop.
Another observation is that whilst the design could be described as an amplifier with noiseless base-emitter feedback , it does not appear to be the same design as the augmented feedback design paper referenced in the Trask article, or used in simple Norton amplifers, because there is no feedback from collector to emitter. Hence the higher gain and relatively modest noise performance, but nevertheless if reports are correct it still seems good enough for the job.
Build the LZ1AQ instead.
It's a better performer and easier to construct, plus I'm not 100% convinced about George's reverse engineered circuit, especially the transformer windings and specifically the configuration on the binocular core. Norton amplifiers bring about lots of other problems and after playing with a lot of different designs, I'm not sure they are worth the effort.
My initial IMD measurements were constrained by my test setup, with my better test rig I measure
LZ1AQ OIP2 +79dBm OIP3+36dBm
Wellgood OIP2 +57dBm OIP3+37dBm (note that the Wellgood is based on an early Wellbrook design and the new ones are better)
The overall performance of all loops, especially on the HF bands is determined by loop size and inductance and amplifier input impedance. This is why most loops 'run out of steam' above about 10MHz.
I've found that contrary to most wisdom, if you can't build a nice 'fat' low inductance loop, it would seem to be better to use an amplifier with a higher value of input impedance (tens of Ohms), in order to improve performance on the HF bands (where the loop inductive reactance dominates the feed point impedance), and compromise a bit on the LF bands, where the natural noise floor tends to be a lot higher (particularly in urban areas) and the very strong Broadcast Stations tend to be problematic anyway.
Some more info on my Active antennas web page
Here's a link I found to a lot of Clifton Labs Norton amplifier circuits (grab them while you can) that may be of interest to you and others on here (use Google translate)
Interestingly for some reason they also produced a higher gain amplifier using four Gali-74's the Z10046A, but I still haven't been able to track down a circuit of the amplifier they produced for the Pixel Loop now sold by DX Engineering (they have withdrawn all Clifton Labs on-line circuit diagrams since they bought them) as the RF-PRO-1B
Send me a private message if you wish.
Martin - G8JNJ
Many thanks for all the info., sounds as if you have built a lot of these loops and noted about the higher Z input needed at higher HF frequencies.
Not had time to finish the loop yet to try the preamp., but if the gain dropoff of the Wellgood from 23dB at 20MHz down to 13dB at 30MHz (measured with 50 ohms in and out) combined with the inductive reactance mismatch is critical, then the Wellgood needs modifying or replacing. BC550s may not be ideal anyway.
I will report back here anyway with results so other readers know how it works with the small design changes and different transistors.
Not saying it is a solution for those who have space or other options but in high noise small city locations it might be a way to actually grab some signals.
I have built one out of 8mm Aluminium tube, junction box(s), binocular Balun, N-Type with external Chinese no-brand 20db LNA.
Here noise can be high and any antenna in the loft just impossible to listen to, e-field just out of the question, this loop is currently an option I can listen to, even if signals are a bit weak.
The image below has excess wiring (+bad solder) for changing baluns, I've yet to optomise it or the LNA setup but it shows promise.
If you life in a noisy location I think it is worth a go.
I have to find some mechanical way to stop the tube rotating in the box so will probably combine the ground with that.
I'm not worried about aesthetics it this point just hoping to prompt feedback from others who might try building the loop and save me time testing options.
With the UK lockdown I've got busier while others seem to have more time.