Keyboard Shortcuts
Likes
Search
LZ1AQ possible changes
|
? Hello everyone, ? a while ago there was a discussion in a german forum where forum member ArnoR proposed modifications to the original LZ1AQ circuit. The modifications seemed plausible to me but as an electronic hobbyist with little knowledge (but a lot of passion!) I had difficulties understanding them. A circuit (and later an updated version) was presented as an improvement over the original LZ1AQ design that i know has been in use for years (and has highly regarded updated versions published in this forum). Since I don’t have the experience to evaluate the design myself, I thought I’d bring it here, where the level of expertise is much higher. My goal is not to argue for or against the changes but simply to understand whether and why they might be beneficial—or if they introduce new drawbacks. Even if there is no benefit, the proposed changes are still interesting from a circuit design perspective. I would appreciate any insights you can share! (I created some LTSpice files for the proposed circuits, they do run but are not refined enough to do a proper circuit simulation. They should be seen as a starting point for experimentation. They are called Arno_V1.asc and Arno_V2.asc in the files section.) ? Here is a summary of the changes:
2. Base Circuit Design Issues
3. Emitter Circuit Design Issues
4. Signal Tapping at Collector Resistors
5. Frequency Response & Component Choice
6. Final Circuit Comparison
? I’d love to hear your thoughts. Could they offer any advantages, or might they introduce unintended issues? Looking forward to learning from your insights! Best regards, Nils Just to be super clear: none of this is my original work, all work was done by ArnoR. I just translated his post to post it here and condensed it for clarity. ? For reference i include the original forum post in translated form: ? Source:
? Since I previously criticized LZ1AQ’s circuit without providing specifics, I now want to briefly address this to avoid the impression of baseless complaining. Let us start at the very end. The output transformer supposedly has a winding inductance of 18?H. With the output resistances of the emitter circuits (220Ω), which generate the output voltage, and the load resistor, the resulting lower cutoff frequency is around 500kHz. This means the inductance is far too low if one intends to amplify cleanly down to the longwave (LW) range. *) The core is specified with ?=1000 and core size R10. According to the Epcos catalog, this core has an AL value of 407, which, with 5 turns, results in an inductance of only 10.2?H. This would place the lower cutoff frequency at about 800kHz. Thus, the lower cutoff frequency is not limited by the input resistance of the circuit and the loop inductance but by the incorrect dimensioning of the output transformer. The two base circuits are powered by separate base voltage supplies. However, the circuit is actually supposed to amplify the differential signal between the emitters. This only works properly if there is no differential voltage between the base connections. Here, this is achieved for AC signals using bypass capacitors. A much more natural approach would be to connect the base terminals directly, thereby constructing a true base-coupled differential amplifier. This would require only a single bias voltage divider, with changes in base currents perfectly canceling each other out. No interference voltage could be coupled between the bases, and there would be no difference in operating points due to resistor tolerances. This results in better symmetry, saves some components and power, and delivers improved performance without any drawbacks compared to separate base circuits. The frequency response remains identical to that of the separate base circuits. At the same time, the lower resistor of the divider could be replaced with an LED, which provides good thermal compensation of the operating point while also serving as an operating indicator. The same mistake as in point 2 has also been made in the emitter circuits. These, too, can be converted into a true differential amplifier without any drawbacks. This again saves several components and improves performance—without even affecting the output drive capability (large-signal behavior). The most serious error, however, is the incorrect signal tapping at the 220Ω collector resistors. The output signal of the base circuits is their collector current, or the voltage across the collector resistors—not the voltage at the collector relative to ground. If, as in LZ1AQ's circuit, the emitter circuits are driven against ground, then the supply voltage and any noise on it appear directly in the signal. Additionally, the operating point becomes highly dependent on the power supply. For these reasons, the circuit can only be operated with a stabilized supply voltage (the 10V regulator). This issue can be easily avoided by using PNP emitter circuits or a PNP differential amplifier. This improves power supply rejection by orders of magnitude compared to the original circuit, eliminating the need for supply voltage stabilization, reducing component count, and increasing output drive capability. The upper cutoff frequency of the original circuit is about 10MHz. It is determined by the effective capacitances at the collector resistance of the base circuits: the Miller capacitance of the emitter circuit, feedback and output capacitance of both the base circuit and the emitter circuit itself. The transistors are no longer suitable for these frequencies or dimensions because their capacitances are too large. According to the author, the circuit is supposed to maintain a flat frequency response up to about 40MHz. However, this is due to the input-side VHF filter introducing a resonance peak in the 10MHz–40MHz range, which compensates for the amplifier’s frequency response roll-off. I consider this an unclean approach. Much better performance can be achieved with more suitable transistors, such as the 2SA1015/2SC1815. These transistors are extremely low-noise, highly linear, have much smaller capacitances, and are very inexpensive. With these transistors, one achieves a significantly higher upper cutoff frequency, a higher slew rate, and a lower input resistance than with 2N2222A, leading to a lower cutoff frequency at the loop. Finally, I present the circuit resulting from the above considerations in the attached images and compare its characteristics with the original circuit. In both cases, the same 1m, 3.4mm AL loop was used, and the VHF input filter was omitted to focus solely on the amplifier’s characteristics. The input signal was kept the same in both setups, slightly into overdrive, to show the maximum output level. The operating currents of all stages are identical in both circuits. Conclusion: Significantly better performance despite much less effort. ? Source:
? Transistors T1/T2 and T5/T6 each form a composite transistor (Sziklai pair) configured as a common-base circuit. The base is clamped (via D1/C1), and the input signal is fed into the composite emitter. This composite transistor has a significantly lower emitter input resistance compared to a single transistor, which is essential since the lower cutoff frequency of the circuit is determined by the relationship: fu=re2πLf_u = \frac{r_e}{2\pi L}fu?=2πLre?? The two composite transistors operate as a base-coupled differential amplifier for the floating magnetic loop, which is connected to the two blocks at the bottom. The output signal from this input differential stage is amplified by the emitter-coupled differential stage T3/T4 and then fed via C8/C9 into a balun, which sums the output signals and provides high common-mode rejection. The resulting 50Ω output is floating (potential-free). C11 and C12 limit the upper frequency response and must be selected according to the desired bandwidth. I have used only C12 with a few picofarads. The input impedance of the circuit is 0.4Ω differentially (i.e., 0.2Ω per side), which allows for a lower cutoff frequency of approximately 20kHz (-3dB) with a 1m loop. The balun is simply a bifilar winding on a toroidal core and is connected like a standard common-mode choke, meaning both winding starts are connected to C8/C9, and the winding ends go to the output. The core must have low losses in the desired frequency range, and the winding inductance should be at least 100?H. ? ? ? |
|
Should you not add my suggestion of a Full Wilson Current Source for each of R1 and R2 of present Everett Sharp circuit.? ( On the other hand, one could try the simple single BJT or FET constant current source firstly......again the THAT CORP 300 series of 4 transistors is a good start....likely they have a SPICE data model available) That can be modelled....anyone out want to try that and see any affect on IP3 best wishes Paul EE VE3PVB
On Sunday, February 9, 2025 at 08:26:29 a.m. EST, Nils via groups.io <nilsp@...> wrote:
? Hello everyone, ? a while ago there was a discussion in a german forum where forum member ArnoR proposed modifications to the original LZ1AQ circuit. The modifications seemed plausible to me but as an electronic hobbyist with little knowledge (but a lot of passion!) I had difficulties understanding them. A circuit (and later an updated version) was presented as an improvement over the original LZ1AQ design that i know has been in use for years (and has highly regarded updated versions published in this forum). Since I don’t have the experience to evaluate the design myself, I thought I’d bring it here, where the level of expertise is much higher. My goal is not to argue for or against the changes but simply to understand whether and why they might be beneficial—or if they introduce new drawbacks. Even if there is no benefit, the proposed changes are still interesting from a circuit design perspective. I would appreciate any insights you can share! (I created some LTSpice files for the proposed circuits, they do run but are not refined enough to do a proper circuit simulation. They should be seen as a starting point for experimentation. They are called Arno_V1.asc and Arno_V2.asc in the files section.) ? Here is a summary of the changes:
2. Base Circuit Design Issues
3. Emitter Circuit Design Issues
4. Signal Tapping at Collector Resistors
5. Frequency Response & Component Choice
6. Final Circuit Comparison
? I’d love to hear your thoughts. Could they offer any advantages, or might they introduce unintended issues? Looking forward to learning from your insights! Best regards, Nils Just to be super clear: none of this is my original work, all work was done by ArnoR. I just translated his post to post it here and condensed it for clarity. ? For reference i include the original forum post in translated form: ? Source:
? Since I previously criticized LZ1AQ’s circuit without providing specifics, I now want to briefly address this to avoid the impression of baseless complaining. Let us start at the very end. The output transformer supposedly has a winding inductance of 18?H. With the output resistances of the emitter circuits (220Ω), which generate the output voltage, and the load resistor, the resulting lower cutoff frequency is around 500kHz. This means the inductance is far too low if one intends to amplify cleanly down to the longwave (LW) range. *) The core is specified with ?=1000 and core size R10. According to the Epcos catalog, this core has an AL value of 407, which, with 5 turns, results in an inductance of only 10.2?H. This would place the lower cutoff frequency at about 800kHz. Thus, the lower cutoff frequency is not limited by the input resistance of the circuit and the loop inductance but by the incorrect dimensioning of the output transformer. The two base circuits are powered by separate base voltage supplies. However, the circuit is actually supposed to amplify the differential signal between the emitters. This only works properly if there is no differential voltage between the base connections. Here, this is achieved for AC signals using bypass capacitors. A much more natural approach would be to connect the base terminals directly, thereby constructing a true base-coupled differential amplifier. This would require only a single bias voltage divider, with changes in base currents perfectly canceling each other out. No interference voltage could be coupled between the bases, and there would be no difference in operating points due to resistor tolerances. This results in better symmetry, saves some components and power, and delivers improved performance without any drawbacks compared to separate base circuits. The frequency response remains identical to that of the separate base circuits. At the same time, the lower resistor of the divider could be replaced with an LED, which provides good thermal compensation of the operating point while also serving as an operating indicator. The same mistake as in point 2 has also been made in the emitter circuits. These, too, can be converted into a true differential amplifier without any drawbacks. This again saves several components and improves performance—without even affecting the output drive capability (large-signal behavior). The most serious error, however, is the incorrect signal tapping at the 220Ω collector resistors. The output signal of the base circuits is their collector current, or the voltage across the collector resistors—not the voltage at the collector relative to ground. If, as in LZ1AQ's circuit, the emitter circuits are driven against ground, then the supply voltage and any noise on it appear directly in the signal. Additionally, the operating point becomes highly dependent on the power supply. For these reasons, the circuit can only be operated with a stabilized supply voltage (the 10V regulator). This issue can be easily avoided by using PNP emitter circuits or a PNP differential amplifier. This improves power supply rejection by orders of magnitude compared to the original circuit, eliminating the need for supply voltage stabilization, reducing component count, and increasing output drive capability. The upper cutoff frequency of the original circuit is about 10MHz. It is determined by the effective capacitances at the collector resistance of the base circuits: the Miller capacitance of the emitter circuit, feedback and output capacitance of both the base circuit and the emitter circuit itself. The transistors are no longer suitable for these frequencies or dimensions because their capacitances are too large. According to the author, the circuit is supposed to maintain a flat frequency response up to about 40MHz. However, this is due to the input-side VHF filter introducing a resonance peak in the 10MHz–40MHz range, which compensates for the amplifier’s frequency response roll-off. I consider this an unclean approach. Much better performance can be achieved with more suitable transistors, such as the 2SA1015/2SC1815. These transistors are extremely low-noise, highly linear, have much smaller capacitances, and are very inexpensive. With these transistors, one achieves a significantly higher upper cutoff frequency, a higher slew rate, and a lower input resistance than with 2N2222A, leading to a lower cutoff frequency at the loop. Finally, I present the circuit resulting from the above considerations in the attached images and compare its characteristics with the original circuit. In both cases, the same 1m, 3.4mm AL loop was used, and the VHF input filter was omitted to focus solely on the amplifier’s characteristics. The input signal was kept the same in both setups, slightly into overdrive, to show the maximum output level. The operating currents of all stages are identical in both circuits. Conclusion: Significantly better performance despite much less effort. ? Source:
? Transistors T1/T2 and T5/T6 each form a composite transistor (Sziklai pair) configured as a common-base circuit. The base is clamped (via D1/C1), and the input signal is fed into the composite emitter. This composite transistor has a significantly lower emitter input resistance compared to a single transistor, which is essential since the lower cutoff frequency of the circuit is determined by the relationship: fu=re2πLf_u = \frac{r_e}{2\pi L}fu?=2πLre?? The two composite transistors operate as a base-coupled differential amplifier for the floating magnetic loop, which is connected to the two blocks at the bottom. The output signal from this input differential stage is amplified by the emitter-coupled differential stage T3/T4 and then fed via C8/C9 into a balun, which sums the output signals and provides high common-mode rejection. The resulting 50Ω output is floating (potential-free). C11 and C12 limit the upper frequency response and must be selected according to the desired bandwidth. I have used only C12 with a few picofarads. The input impedance of the circuit is 0.4Ω differentially (i.e., 0.2Ω per side), which allows for a lower cutoff frequency of approximately 20kHz (-3dB) with a 1m loop. The balun is simply a bifilar winding on a toroidal core and is connected like a standard common-mode choke, meaning both winding starts are connected to C8/C9, and the winding ends go to the output. The core must have low losses in the desired frequency range, and the winding inductance should be at least 100?H. ? ? ? |
开云体育
I would disagree with
calling either this circuit or the LZ1AQ circuit differential.
Both are two separate amplifier chains fed with a differential
signal. The only differential coupling in the amplifier is the
output transformer. Doesn't ANY diode
generate noise?? The green LED provides a stable voltage to
protect against power supply variations, but putting a noise
source on the base just feels wrong.? John Kolb?? KK6IL On 2/9/2025 5:26 AM, Nils via groups.io
wrote:
|
|
I suspect that ArnoR has not yet tested an original LZ1AQ amplifier and he did not evaluate it's performance on the bench nor in the field. Some of his judgement seems to be based on simulations and his assumptions, some of which are obviously inaccurate, but some of his suggestions are worth thinking about, e.g. replace the 2N2222A with better performing RF-transistors. ArnoR may be very experienced in analog transistor circuit design. As far as i know, he himself did not yet provide measurement plots proving the achievements of his proposed design compared to the performance of the original LZ1AQ amplifier.
?
For example claim 1:
?
Output Transformer Issues
?
This assumption, IMHO, is ridiculous. The output transformer has not a "too low" inductivity of 10.2 ?H. I dont know where he takes this value from? The push-pull output transformer is wound on a high-permeability ferrite core. The open circuit incuctance of the transformer is likewise in the range of several hundreds of ?H. If he had measured this circuit on the lab bench, he would have known that the frequency range of an original LZ1AQ amplifier goes down much much lower than 800 kHz. In the files section there are many frequency plots which prove that.
?
regards
Fred
|
|
On Mon, Feb 10, 2025 at 05:43 AM, John Kolb wrote:
Doesn't ANY diode generate noise?? Reverse biased diodes in breakdown like Avalanche and Zener-diodes generate noise, a normal diode does it nor or much less. Diode based transistor biasing is frequently used and proven in audio amplifiers.
?
regards, Fred |
|
On Mon, Feb 10, 2025 at 07:19 AM, Fred M wrote:
I dont know where he takes this value from? I think he ist referring to the original schematic on the LZ1AQ website. The transformer there ist trifilar with 18uH each (5 trifilar turns on a toroidal core, u=1000, 10x6x4 mm). He probably recalculated the inductance.
?
Best regards,
Nils |
|
Hi Nils,
?
On my Improved LZ1AQ design the inductance @100Hz is 610uH on each winding and with the two in series it is 2,395uH. The gain from 100kHz to 20MHz is 30dB? and is 27dB at 30MHz. I don't Know how you can improve on that? Also I am using the lowest NF transistors you can find, which is a 1dB NF.
?
Everett N4CY
In a message dated 2/10/2025 6:41:48 AM Central Standard Time, nilsp@... writes: ?
|
|
On Mon, Feb 10, 2025 at 12:41 PM, Nils wrote:
The transformer there ist trifilar with 18uH each (5 trifilar turns on a toroidal core, u=1000, 10x6x4 mm). He probably recalculated the inductance. In the older original LZ1AQ AAA-1B that i own, the transformer is on a 10 mm diameter blue coated Siemens/TDK Sifferit N30 toroid core, ?i = 4300.
?
?
5 turns wound on this core would give an inductivitiy of approx. 45 ?H. The lower -3 dB cutoff frequency of a transformer is determined by the point where the transformer's primary inductance has the same impedance as the impedance driving it. Let's assume 50 Ohms. That results in a lower cutoff frequency of roughly 170 kHz. Below the cutoff frequency the transformer behaves like a low pass filter with a slope dropping with approx. 6 dB / octave. I dont know, wether the newer versions now have a different output push-pull transformer.
?
If i did a mistake in my calculation please correct me.
?
regards
Fred
?
?
?
? |
|
Mike,
toggle quoted message
Show quoted text
It is easy to increase the current, which would improve the IMD, but it will increase the noise. Everett On Monday, February 10, 2025, 5:59 PM, Mike M <groups@...> wrote:
|
|
On Sun, Feb 9, 2025 at 01:26 PM, Nils wrote:
Hi all,
i went through this german "mikrocontroller forum" thread from August 2021, where the threadopener ArnoR claimes, that the LZ1AQ circuit is "much too bad" and proposes a long list of modifications to improve the - in his view - flawed design. (See Nils' opening message of this topic).
?
ArnoR seems to be quite experienced in discrete transistor circuit design, but not explicitly experienced in RF. The? language he uses during the forum discussion reads, even by german cultural language standards, annoyingly rude and self absorbed. He expects to be encouraged in his views and opinions and seems not really interested in advise or a true discussion of technical facts others than his. The moment of truth comes in post 15.08.2021 18:13:
?
The poster "Heiner" is asking the question: "Are your ‘optimisations’ of the LZ1AQ circuit merely based on simulations, or measured in comparison and compared in real in practice under identical conditions?"
?
ArnoR's response: "These are simulations, just like LZ1AQ, whose data also originate only from simulations. And two circuits,? simulated under the same conditions are comparable, even if the models are not 100% correct.......The circuit simulator is an excellent tool for designing such things. Due to the lack of measurement equipment, however, I can only judge subjectively."
?
Let me conclude: The design flaws alleged by ArnoR with respect to the very early circuit diagram published by LZ1AQ, are solely based on his assumptions and simulations. He insinuates that LZ1AQ's experiences are also based on simulations only. ArnoR has obviously neither built nor tested or compared the amplifiers, nor had he the basic RF-test equipment available to do this, respectively big skills in RF-testing or practical dx-ing!?
?
It seems, ArnoR's remarks have remained an unvalidated theoretical exercise, I have not yet found any information that his proposed amplifier circuit was built or tested by anyone later.?
?
regards, fred
|
|
I worked for a company that had among other products a simulator. The architect was somewhat of a curmudgeon but really knew his stuff. He really drove home the idea that you need good measurements to validate your simulations. The best way to enrage him was to show him the all-too-common customer question "why does your simulator give a different answer than this other simulator?".? A much better question would be "why does your simulator not match these good measurements?". There are so many ways that different simulators can vary even with the same models that it is never a fun exercise to determine why they differ.
?
So simulation-based arguments may be interesting, but unless there is the built and tested prototype it is hard to know whether the simulations represent reality or not.
?
--
=================================================================== Mike M |
|
Simulations do have their place, and they allow us to quickly try new things, without letting out the magic smoke.
?
However, so many graduates these days are trained using simulation tools, without ever touching actual components, so it often comes as a bit of a shock to them, when they have their first industry placement, and have to start actually making things.
?
The first challenge is to get them to be able to recognise parts, and the second is in acknowledging that there are various physical constraints that also have to be taken into consideration. Such as power dissipation and cooling, and in the case of RF, screening, stray coupling, including unaccounted for inductance and capacitance, which all need to be taken into account.
?
It always used to amuse me when dealing with one particular make of broadcast transmitter, where individual sections had obviously been designed by graduate engineers, and the feedback around operational amplifier stages used non-standard value, 1% tolerance parts, even though they were intended to be simple, unity gain buffer stages.
?
This should have really been spotted and sorted out during production engineering, but for some reason it never did.
?
Regards,
?
Martin
?
?
On Fri, Feb 14, 2025 at 03:02 PM, Mike M wrote:
So simulation-based arguments may be interesting, but unless there is the built and tested prototype it is hard to know whether the simulations represent reality or not. |
|
On Fri, Feb 14, 2025 at 08:33 AM, Martin - Southwest UK wrote:
It always used to amuse me when dealing with one particular make of broadcast transmitter, where individual sections had obviously been designed by graduate engineers, and the feedback around operational amplifier stages used non-standard value, 1% tolerance parts, even though they were intended to be simple, unity gain buffer stages.This example is more about the cost optimization, which is another side of the real life engineering. It applies to hobby projects as well. A hobby design becomes popular only when it balances the performance with cost (and availability of parts). LZ1AQ loop amp is a nice example. But suggesting 2SC1815 for use in it is a step too far in the cost reduction direction :) ?
73, Mike AF7KR |
|
The other issue that many loop? experts dont want to accept is that for a the typical 1 metre? to 2 metre loop size? the antenna factors cant be improved to the point where it drops the inherent amplifier or system noise floor.
?
If the ultimate objective is a better signal to noise ration then a better antenna design should be pursued rather than making miniscule changes in loop amplifier design that does not change? noise or floor antenna factor by much.? Then again I am not surprised because even the likes of the FCC, OFCOMM and several other regulatory agencies use measuring loop antennas whose? noise floor is higher than the noise that they are trying to measure.? They just dont want to get it!
?
The only thing that will improve antenna factor? to get a better signal noise ratio for a 1? to 2 metre loop size is to implement a tuning/tracking system that gives anything from 6 to 10 db improvement. The only professional company that "gets it" is R&S whose model? HM525 loop has a tuned or amplified mode that results in a substantial signal to noise floor improvement that allows noise measurement? or signal reception close to ITU rural noise levels.
?
So I question the value of pursuing the miniscule improvements in pursuing? minor amplifier and transistor improvements beyond what is necessary for the antennas antenna factor that it will be used in. It seems to be an exercise in futility when the majority want to use a small 1 to 1.5m loop diameter.
?
When someone develops a loop antenna amplifier and small loop? antenna that improves 10db over the existing 1 metre designs I will take notice. That is not to say that all this research and? endeavour is not technically interesting? because it does advance everyone's technical understanding which is great. Ultimately it comes down to antenna factor that nobody wants to talk about!
?
Henry |
|
Before considering changes, the objectives that you want to achieve should be clearly defined. For example
?
a) improved sensitivity/noise figure
b) improved gain / gain flatness
c) improved frequency range
d) improved linearity / intermodulation performance / IP2, IP3
e) lower/higher/tracking input-Z
f) lower/higher supply voltage/supply current/power consuption
g) minimum Common Mode Rejection Ratio over frequency h) lower costs
i) common availability of quality components
j) ESD hardening
k) environmental conditions
k) else . . .
?
Some of these requirements compete with others, some come with a price. Some of them ar very specific or are of limited relevance for general receive purposes. For example, an exeptionally good CMRR, which is achieved over a certain frequency range only may have no practical advantage, because the overall antenna CMRR is limited by external factors. IMHO, implementing changes just for the sake of change makes no sense. It should be known what goals you want to achieve with it.
?
regards,? Fred
? |
|
Hi All,
?
Very interesting thread.
I looked at the circuit introduced. I felt nostalgia for the 2SC1815 and 2SA1015. They like staple foods for us from the 1980s until around 2010.
The 2SA1015 was surprisingly low noise and was used for Rx Preamp in the then hit CB Band Amplifire. I hadn't used a parts with legs in years, so I wanted to build one.
I wrote about it on my blog. It is in Japanese, but I think it makes sense with machine translation.
?
73, Hisami 7L4IOU
?
|
Everett N4CY