TI Bulk Acoustic Wave (BAW) clock oscillator

It's a fabulous concept, putting the resonator on chip. That could be done in uPs and FPGAs and tiny clock chips.

Jitter is fabulous.

But the stability is worse than a good XO.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

That makes sense. Their target is jitter.

What is your example of a "good XO" by part number?

One can use a PLL topology to use the stable XO to stabilize the BAW oscillator, yielding a good mix of stability and jitter.

What is the rough requirement?

Joe

From:

This is nowhere near a Rubidium clock!

'Acoustic (mechanical) resonators' are as old as the world!

Many satellite LNBs (low noise down converters fron GHz to IF frequencies) use those:

But for reception of for example single sideband signals those are not sufficiently stable, I have used the 10 MHz Rubidium oscillator reference to create a stable reference for LNBs that do have a crystal controlled PLL to make the 10 GHz GHz.

Most FPGAs and many uPs have an on-chip oscillator. They are maybe

We use the RP2040 cpu chip. It uses its internal clock for boot/bios startup. It is "guaranteed to operate between 1.8 MHz and 12 MHz."

We use an external 12 MHz XO after bootup. And another one, 25 MHz, for the Ethernet chip.

Oops

A PPM-class on-chip BAW oscillator is revolutionary.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Am 03.07.26 um 18:37 schrieb Jan Panteltje:

These are dielectric resonators. They have nothing to do with acoustic waves. Simply a blob of high eps-r material.

I found a requirement of sorts in the "cheap rubidium" thread: "20 PPB long-term stability" followed by the comment that "a good TXCO or OCXO is about 1 PPB per day".

I have a few reactions:

First, how long is "long-term" in units of time? This value is the primary control of what kind of physics must be used.

Second, a big difference between a TCXO and a OCXO is the handling of HVAC cycling. Typically, inhabited facilities have a 15-minute control cycle, and one will see the HVAC cycle peak at around 1000 seconds in TCXO output, but not much in OCXO output.

Third, how important is jitter in your application?

Joe

I just wanna be right, so: From:

Fundamental Mode Dielectric Resonator Oscillators In this type, the dielectric resonator operates in its fundamental resonant mode. The oscillator generates a signal at a frequency corresponding to the fundamental resonance of the dielectric material. This type is relatively simple and commonly used in various frequency control and timing applications.

Higher Order Mode Dielectric Resonator Oscillators Unlike the fundamental mode DRO, higher order mode DROs utilize higher order resonant modes of the dielectric material. This allows for the generation of higher frequency signals. RF Engineers can select specific modes to achieve the desired frequency output, providing flexibility in oscillator design.

Dielectric Resonator Filters with Oscillation This type combines the functionality of a dielectric resonator filter and an oscillator in a single device. They operate at the resonant frequency of the dielectric material, serving both as a filter and an oscillator. This integration simplifies the overall system design and can be advantageous in certain applications.

Dielectric Resonator Stabilized Oscillators This type incorporates dielectric resonators to improve the frequency stability of traditional oscillators, such as Gunn diode oscillators. The dielectric resonator stabilizes the oscillation frequency, reducing phase noise and enhancing the overall performance of the oscillator.

Dielectric Resonator Varactor Oscillators This type utilizes varactor diodes in conjunction with dielectric resonators to achieve tunability in the generated frequency. By varying the bias voltage applied to the varactor diodes, the capacitance changes, enabling frequency tuning. This type of DRO is suitable for applications requiring frequency agility.

Phase-Locked Loop (PLL) based Dielectric Resonator Oscillators Some DROs are integrated into phase-locked loop systems. In such configurations, the DRO serves as the frequency reference, and the PLL circuit ensures precise control and synchronization of the oscillator frequency. This type is often used in applications where frequency accuracy and stability are critical. DRO based PLLs using high-quality reference oscillators combine very low phase noise of such references at low offset frequencies with low phase noise of the DROs at higher frequency offsets.

Surface Acoustic Wave (SAW) Resonator Dielectric Resonator Oscillators SAW resonators, which utilize piezoelectric materials, can also be used as resonant elements in DROs. These DROs benefit from the unique properties of surface acoustic waves, offering advantages such as small size and low power consumption.

Multi-Resonator Dielectric Resonator Oscillators This type of DRO utilizes multiple dielectric resonators to achieve stable oscillations across a broader frequency range. By selecting different resonators, each tuned to a specific frequency, engineers can design oscillators capable of operating in multiple frequency bands.

What are the Advantages of Using Dielectric Resonator Oscillators (DROs)? Frequency Stability One of the primary advantages of DROs is their exceptional frequency stability. The resonance frequency is primarily determined by the physical dimensions and dielectric properties of the resonator, leading to minimal frequency drift over time and temperature variations.

Low Phase Noise Phase noise is a critical parameter in many frequency control and timing applications. DROs typically exhibit low phase noise characteristics, making them suitable for radio frequency and microwave systems requiring high signal purity and minimal interference.

So, it's all resonating...

Did you notice the difference in size of the 2 white round things for the different GHz frequencies? 10 GHz corresponds to a wavelength of about 0.03 meters. (3 cm ) looks like 1/4 of that, have not measured it, almost like a cavity?

Anyways it is all electrically induced RF vibration.

The soldering needs some work!

Yes, smaller and smaller..

It does not.

We theorized that we could use a single 25 MHz XO on our RP2040 products, and maybe we could have, but running the Ethernet and the code-load USB interface from the same XO turned out to be a huge tangle, so we kluged in a second oscillator. Since I'm the best tiny-part solder-person in my shop, I hacked them.

Fortunately, this was on a proto board, and that's what proto boards are for, to try things.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

One of my hobbies is designing super-stable instant-start oscillators. All our current products use LC oscillators, which have evolved over about six generations.

I did try some based on coaxial ceramic resonators.

CCRs are common in RF systems. Q's are crazy high (thousands) and tempcos are crazy low (PPMs.)

My triggered CCR oscillators worked, but low frequency ("low" being

CCRs do not vibrate. They are bits of coax with very stable high-K dielectrics.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Apart from LC filters, my first encounter with narrow band filters was a cystal filter. Specifically the XF9B 9 MHz crystal filter in the transmitter I build in the early sixties

Wound hundreds of coils for LC filters ever since...

Trion T20F256 FPGA evaluation board:

I never used a Trion FPGA, any good? I can use my Raspberry as a programmabe oscillator, or FM transmitter:

Wrote this: Raspberry Pi as signal generator from 150 kHz to 500 MHz:

I looked into TI's BAW technology, which is of course heavily patented. There are many patents, mostly discussing the foundry processes to make chips cheaply enough.

But a recent one on an improvement gives some background. It is US

So, one could implement a suitable crystal-oscillator oven using Peltier-junction units (which can heat or cool) with the BAW oscillator within. A single-oven design may suffice.

Joe

Yes. They are cheap and simple, and the software is free and simple. None of the licensing nonsense like Xilinx or Altera.

That FPGA runs a 250 MHz DDS and a delay generator core and four pseudo-random shift registers and pulls 25 mA core current.

The combination of RP2040 and T20 is great.

The RP2040 has a couple of GHz PLLs. Some lunatic overclocked one to 1 GHz in liquid nitrogen.

The DDS works up to 50 MHz. We synthesize an octave, 12 to 25, and divide down from there. That keeps the jittter down.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Thank you, maybe if I need some similar stuff I will try that evaluation board.

LOL, yes, that's cold.. The 250 years US festivity crowd would like that. I did read they were chased way by a thunder warning.. God speaks?

What's wrong with ECL dividers? They are current hogs, but so is anything that run that fast. Presumably there are programmable logic parts around now that can also run fast enough.

The 5R impedance is a bit of a pain, but you might look at a transmission-line transformer to step it up to something easier. They only do integer ratios.

Matick R E 1968 "Transmission-Line pulse transformers - theory and applications" Proc. IEEE. 56 pages 46-72

Which vibrate very rapidly with an absolutely tiny amplitude.

Bipolar logic - rtl, dtl, ttl, ecl - is SO last millenium.

An ECL flop is a million times as big as a flop inside an FPGA, and

We rarely use bipolar transistors any more either.

And only transform AC.

Sound waves from local pontification.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

before the latest version(which you can still use) Xilinx had no license nonsense as long as you weren't using the huge parts

tried the the newer rp2350 that has twice the ram, faster M33 cores, and hardware single precision FPU?

afaiu it easily runs at 400MHz at 1.3V