ChatGPT on diode ring mixers

[allenc]

So I thought I'd hit ChatGPT with a question about LO signals and diode ring mixers.  A frequent topic of discussion about this whether to use sine or square wave LO input.  I've seen both answers offered in various interweb discussions.  Here is what I got from ChatGPT.  Discuss:

> Should the local oscillator signal for a diode ring mixer be a sine wave or square wave?

The local oscillator signal for a diode ring mixer should ideally be a square wave. This is because a square wave has odd harmonics, which are required for the mixing process to occur in a diode ring mixer. The odd harmonics of the square wave generate the local oscillator signal's necessary frequency components, which are used to mix with the input signal.

When a square wave is applied to the diode ring mixer, it generates a series of odd harmonics, which are used for the mixing process. These odd harmonics combine with the input signal to produce sum and difference frequencies, which are used for signal processing.

In contrast, a sine wave does not have odd harmonics, which means it is not suitable for use as the local oscillator signal in a diode ring mixer. While a sine wave can be used in some types of mixers, such as the Gilbert cell mixer, it is not suitable for use in a diode ring mixer.

> How do odd harmonics facilitate the mixing process in a diode ring mixer?

Odd harmonics play a crucial role in the mixing process of a diode ring mixer. When a square wave is applied as the local oscillator signal in a diode ring mixer, it generates a series of odd harmonics (i.e., harmonics with frequencies that are odd multiples of the fundamental frequency). These odd harmonics are used for the mixing process, which involves multiplying the input signal with the local oscillator signal to produce sum and difference frequencies.

The mixing process in a diode ring mixer occurs because of the non-linear properties of the diodes in the ring. When the input signal is applied to the diode ring, it is rectified by the diodes, which causes it to generate harmonics at frequencies that are multiples of the input frequency. The local oscillator signal, which is also a square wave with odd harmonics, mixes with these harmonics generated by the input signal. This mixing process produces sum and difference frequencies that can be extracted from the output of the mixer.

The odd harmonics of the local oscillator signal are essential for the mixing process because they provide the required frequency components for the multiplication process to occur. If the local oscillator signal were a pure sine wave, it would not contain odd harmonics, and therefore, it would not be suitable for use in a diode ring mixer. The odd harmonics generated by the square wave local oscillator signal provide the necessary frequency components for the mixing process to occur and allow for efficient frequency conversion in the diode ring mixer.

[antenna]

Sounds like a fancy way of saying square waves are preffered because it forward biases the diodes faster allowing the phase change to take as little time as possible.  ChatGPT is using the effect to explain the cause in my opinion..

[antenna]

And this is why I asked the question on a youtube video that, like many other places, had showed the IF connected to the diodes and the LO on the other side of the transformer.  I figured having the LO signal going through the transformer would round the edges a little and slow the switching speed and that the LO should be the signal at the diodes, but I got crickets.  No response from anyone, not even from the creator.  In fact, it says no comments, but shows mine only when logged in.  Comments must be disabled by the creator.

[Odysseus]

Are you suggesting to use the as-drawn output port as the the LO connection? The three ports of a diode ring mixer aren't interchangeable.

[antenna]

Im not suggesting, I am asking.  I dont understand why one would apply the square wave signal to a connection isolated by a transformer.  i thought the LO (what I assume is the square wave) to the connection labeled output as those are the connections at the diode.  I probably misunderstand it.

[Odysseus]

To your point, the LO port is often connected using a wide bandwidth transmission line transformer, which can be an improvement over the inherent bandwidth limitations of a conventional transformer configuration. The schematic ends up looking like the following attachment:

Regarding the original post, Chat GPT sounds to be talking in circles and is pretty far off the mark. Either drive type can work, with pros and cons of each. For example, a square wave may improve the conversion loss by minimizing the off-time of the diode switches, but may induce higher levels of LO feed-through.

[vk6zgo]

To your point, the LO port is often connected using a wide bandwidth transmission line transformer, which can be an improvement over the inherent bandwidth limitations of a conventional transformer configuration. The schematic ends up looking like the following attachment:

Regarding the original post, Chat GPT sounds to be talking in circles and is pretty far off the mark. Either drive type can work, with pros and cons of each. For example, a square wave may improve the conversion loss by minimizing the off-time of the diode switches, but may induce higher levels of LO feed-through.

Sounds pretty much "par for the course".
When I have asked it technical questions, the answer is rarely valid, resembling the muddled stuff we often see from a "noob".

[T3sl4co1l]

ChatGPT is using the effect to explain the cause in my opinion..

It isn't explaining anything; it's just pasting together words and phrases it's seen.  Presumably it's seen enough articles on the topic to have some preference for one of the possibilities -- but it doesn't have any reasoning, inference, any semantic or factual knowledge underlying its otherwise excellent grasp of language.

Example: true insight is realizing that the distinction is soft (a matter of definition, not a physical phenomenon), and it goes on a continuum from small-signal sine to heavily overdriven sine (which ends up clipped by the diodes to a reasonable approximation of a square wave).  More drive means less mixer loss, but more harmonics.  You generally want enough level so that LO dominates at most signal levels, but not so much that it overheats, or that harmonics / higher order mixing products are excessive.  Pick your compromise.  That is all.

Tim

[ejeffrey]

By the way: if the goal is to discuss chatgpt and how it answers questions, there is a separate message board for that in the computing section.  Overall I'd say the answer given is basically muddled garbage.

Mixers are all basically commuted switches in one form or another.  In a diode mixer the signal and the switch control (LO) both go through the diodes the same way.  So it is important to hard-switch the diodes otherwise the signal itself will be partially controlling the switch during the transition and you will get intermodulation of the signal with itself.

You can hard switch with a square wave, or just a large amplitude sine wave

It's a side effect of this behavior that you get all the odd harmonics of the LO (including mixing products of those harmonics with the signal) but those are generally far out of band and easily filtered.

[Roger Need]

I asked the new Bing AI in Microsoft Edge the same question and got this answer.

Should the local oscillator signal for a diode ring mixer be a sine wave or square wave? Provide references

A diode ring mixer is a type of balanced modulator that produces double-sideband suppressed-carrier (DSBSC) signals by mixing a single frequency carrier with a modulating signal. It consists of four diodes arranged in a ring and input and output transformers. The carrier signal is applied at the center taps of the transformers and the modulating signal at the primary of the input transformer. The output is measured at the secondary of the output transformer.

One of the questions that arises when designing or using a diode ring mixer is whether the local oscillator (LO) signal that drives the carrier input should be a sine wave or a square wave. The answer depends on several factors, such as the desired performance, the available components, and the trade-offs involved.

A sine wave LO has the advantage of being a pure tone that does not contain any harmonics or spurious frequencies. This means that it will only produce the desired sum and difference frequencies at the output, and not any unwanted intermodulation products or distortion. A sine wave LO also has a lower peak-to-peak voltage than a square wave LO with the same average power, which reduces the stress on the diodes and increases their lifetime.

However, a sine wave LO also has some drawbacks. It requires a more complex and expensive circuit to generate, such as an oscillator followed by a filter or an amplifier. It also has a lower switching speed than a square wave LO, which means that it will not fully turn on and off the diodes in each half-cycle. This results in some residual carrier leakage at the output, which reduces the suppression ratio and increases the noise floor.

A square wave LO, on the other hand, has the opposite characteristics. It has a simple and cheap circuit to generate, such as an oscillator followed by a buffer or a driver. It has a high switching speed that fully turns on and off the diodes in each half-cycle, which eliminates any carrier leakage and improves the suppression ratio and noise floor.

However, a square wave LO also has some disadvantages. It contains many harmonics and spurious frequencies that will also mix with the modulating signal and produce unwanted intermodulation products and distortion at the output. A square wave LO also has a higher peak-to-peak voltage than a sine wave LO with the same average power, which increases the stress on the diodes and decreases their lifetime.

Therefore, there is no definitive answer to whether a sine wave or a square wave LO is better for a diode ring mixer. It depends on the specific application and design requirements. In general, a sine wave LO is preferred for high-performance applications that require low distortion, low noise, and high suppression ratio. A square wave LO is preferred for low-cost applications that require simple circuitry, high switching speed, and high efficiency.

Some references that discuss this topic in more detail are:

- [6.3: Single-Ended, Balanced, and Double Balanced Mixers - Engineering LibreTexts](https://eng.libretexts.org/Bookshelves/Electrical_Engineering/Electronics/Microwave_and_RF_Design_IV%3A_Modules_(Steer)/06%3A_Mixer_and_Source_Modules/6.03%3A_Single-Ended_Balanced_and_Double_Balanced_Mixers)
- [Activity: Diode Ring Modulator - ADALM2000 - Analog Devices](https://wiki.analog.com/university/courses/electronics/electronics_lab_diode_ring_modulator)
- [Taking the mystery out of diode double-balanced mixers](https://www.robkalmeijer.nl/techniek/electronica/radiotechniek/hambladen/qst/1993/12/page32/)

[DaJMasta]

It isn't explaining anything; it's just pasting together words and phrases it's seen.  Presumably it's seen enough articles on the topic to have some preference for one of the possibilities -- but it doesn't have any reasoning, inference, any semantic or factual knowledge underlying its otherwise excellent grasp of language.

This.  There is no understanding in these algorithms, it's effectively free association that has been checked millions of times against existing content.  It's not that what it's going to tell you is wrong, its that it doesn't reflect understanding or skill - it's just data pulled from a pool of random sources of every level of quality.

I have a linguist friend who told me she though ChatGPT and AIs are basically the ultimate bullshitters.  They don't know what they are saying, and they are not designed to, but they "learn" to say it as well as possible.  Her example was to ask it why the ocean was blue - and its answer was the opposite of the actual reason (it said that water absorbed blue light).

Don't use any current AI tool to teach you a topic, it doesn't know even if it has access to great sources, but there can be something to be gained from feeding it factual information and seeing how it frames or explains it.  There are absolutely things it's good at, but understanding or deciding what's true or not are not them.

[RoGeorge]

ChatGPT gets old fast ain't it?

Everybody gets fooled at first, or at least I was, but ChatGPT is not what it looks like at the first look.  Those are aggregated words - not aggregated knowledge.  ChatGPT returns a phrase from words that are usually lumping together in the tons of text that was feed to it at training.  The result may seem quite impressive, though don't take those ChatGPT words as correct, or as meaningful.

Dave made a dedicated section to contain ChatGPT flooding all the forum:
https://www.eevblog.com/forum/chatgptai/