RF-Frontend  Yaesu FT-221R / FT-225RD

After obtaining my radio amateur novice license in 1977, a transceiver was of course immediately needed. It became the Yaesu FT-221 R. A 2-meter all-mode transceiver that, with its specifications, mechanical construction, and modular design, stood miles above everything else available at that time. A beautiful device. However, I had to part with it quickly. Since my true passion was on the shortwave bands, I immediately started working on CW to obtain my novice license. A Kenwood TS-430 S came into play, and the FT-221 R was sacrificed.

But as it goes with true love, the regret remained.

Forty years later - and after many searches on eBay and Marktplaats - it finally happened. For a price of only €85, I could get my hands on a truly impeccable FT-221‌‌R, along with a speaker in the same style that I had never seen before. A black vintage SP-101? According to the seller, they had both spent 40 years under a blanket. And indeed, it looked better than the one I had parted with back then.

The FT-221 R in mint condition with a matching speaker
The design with its eleven plug-in cards makes it particularly possible to overhaul parts of the device without affecting the whole or making irreversible changes. Although even by current standards of reasonable quality, the RF reception board is the first to be considered b ecause the sensitivity of the receiver leaves something to be desired.

The Yaesu PB-1456 'RX RF board' and its replicas

The original Yaesu board

In the photo above is the original RX RF Unit from my FT-221 R. According to specifications, the sensitivity of the receiver is 0.5uV @ 10dB S/N for SSB and CW, and 0.75uV @ 20dB QS for FM. The moderate sensitivity is partly attributed to the mediocre noise figure of the MOSFET at the input (a 3SK51). But also, four Varicaps that are used to tune the narrow band filters on the board with the band selection switch contribute significantly to the noise. The entire setup also does not excel in terms of strong-signal properties. Forty years ago, different requirements could be imposed on it than are necessary today.

The (G3SEK) MuTek-board

I still remembered that there was much praise for the so-called Mutek board at that time. So, first, I looked on the internet to see what could still be found about it. Already in 1983, Ian White (G3SEK) designed a high-quality replacement RF reception board. This was then marketed by Mutek Ltd in Oxford, UK under the name "RPCB 144ub FT221/225 front-end board." Ian gave the following reasons for his initiative:

The original rf board fitted by Yaesu suffers from several deficiencies. The chief of these are a distinct deafness (noise figures of 8 - 10dB are nou unsual in unmodified tranceivers!) and a considerable susceptability to strong- signal overload problems. Fitting a preamplifier can help wit the first problem but at the expense of dynamic range."

To address these issues, the input stage of his board is equipped with a low-noise dual-gate MOSFET (a 3SK88 with a noise figure of approximately 1dB). In addition, there is a Chebyshev band-pass filter that covers the entire 4 MHz wide range, eliminating the need for narrow-band tuning with Varicaps. Furthermore, there is a Minicircuits diode ring mixer (an SBL-1 with a 1dB compression point of +1dBm), a number of 10.7 MHz amplifiers operating at a slightly higher voltage than the original board, and a 6-pole crystal filter. All in all, a significant improvement over the original design. The only problem is that the board was only on the market for a limited number of years. It is now a collector's item that people are still looking for. Not surprisingly, new designs have been created.

I have found two of them.

The IZ4BEH-board

In 2008, an article appeared in the Italian magazine Notiziario VUSHF (July 2008 page 2 et seq.) by Roberto IZ4BEH, where the main modification compared to the Mutek front-end was to replace the MOSFET at the input with an ATF54143 GaAs pHEMPT from Avago and to increase the level of the local oscillator to apply 17dBm and 23dB LO mixers (such as the SBL-1H, SRA-1H, and RAY-1).

The IZ1DYE-board

Another design came from Michele IZ1DYE in 2010 which has many similarities with the previous one. For unclear reasons, an additional ERA-5 MMIC has been included between the LNA and mixer. Furthermore, the board is predominantly equipped with SMD components.

The PA3AJR-board that is presented here

As far as I can tell, neither of these successors to the Mutek board has ever been commercially available, and no alternatives have appeared. So, I decided to build something myself! Creating a new design from scratch is not really an issue with the current PCB design and production capabilities. It also provides an opportunity to make further technical improvements, choose common components, and use some items that are already on the shelf. This has led to the board in the photo next to it, and I must also mention that I have no intention of bringing it to market as a commercial product. But if I can assist someone with words and deeds, I am always willing to do so.

Design Principles

Below are some considerations and points of attention that played a role in the construction.

  • No irreversible changes to the transceiver itself
    Given the original state of the set being preserved for 40 years and the modular design inviting changes, completely swapping the board is an excellent option to keep the device fully intact. One modification I have implemented, unrelated to this RX RF Unit, is a modification in the PLL unit as described by SM5BSZ. This intervention aims to limit the phase noise of the LO signal, addressing a potential bottleneck in reception quality with a relatively minor adjustment.
  • No surprises in the execution
    With a prototype and a small initial production run, along with long lead times for acquiring parts, it is crucial to avoid overly critical components in the design and keep things predictable. Preferably, no trial and error on the final product!
  • It must remain affordable
    The entire design (including board production) should not cost more than a few tens of euros. The goal is to bring Yaesu's museum piece into the modern era, but I don't envision returning to enthusiastic use of the 2-meter band. It needs to remain reasonably proportional. The joy is in the design process, not necessarily in the usage.
  • Suitable for various crystal filters
    The board has enough space for a "full size" crystal filter commonly found in many VHF/UHF mobile radios (from manufacturers like KVG, ITT, Dantronic, Salford) and readily available to me. Additionally, there is now a range of decent and affordable 6-, 8-, or 10-pole filters in a 23 x 12mm SC-4 housing (and 13.4 x 5mm pitch), such as those from Toyocom (10M5F3 50) en ECS (ECS-10.7-D ). Prices on eBay range from €5 to €15. Merging the footprints of both types into one combined option is a straightforward task. Furthermore, in line with the original board's minimalist approach, incorporating two monolithic filters in HC-49T housing (e.g., 10M15A at €2.50 for 5 pieces) presents three physical options.
  • Provisions for testing and tuning

    Test points have been placed around the mixer - on the top side of the board - to measure or inject RF, IF, and LO signals. To facilitate adjustments outside the device, a separate connector is used as a test jig (Card Edge Connector Ra805 series, 18 pins, 3.96mm pitch). Once inserted into the device, the board is less accessible. To perform measurements and adjustments in this scenario, a design for a 2 x 18-pin Extender Board was created concurrently with the production of the reception board. The double-row pins make this extender suitable for maintaining all other boards in the transceiver. As the extender can also be useful for many other devices, the details and production files are available in a separate topic.

    The test jig and the extender board."
    The FT-221 R with the RX RF Unit on the extender board
  • Attention to good availability in small quantities of components

    The 10.7MHz transformers - often from TOKO - are readily available on eBay and AliExpress, usually identifiable by a pink, purple, or orange tuning core. I typically remove the capacitor at the bottom and reintroduce it as an external capacitor on the board. Removing a coil form with all its pins generally does not improve the board's appearance. For the 145MHz RF and LO coil forms , I used 'Neosid 10 V 1 Coil Assemblies.' Whenever possible, SMD components have been used.

Don't be overly concerned about soldering SMDs. It's not as difficult as it seems. Tin one pad first with a fine soldering iron. Then use tweezers to hold the component in place and reflow the solder. Finally, solder the remaining pads. A good magnifying glass and a flux pen make it all much easier. A hot air soldering station is very convenient, but for removing resistors and capacitors, I prefer to use a second soldering iron. Resistors and capacitors in 603 or 805 are available in large quantities, cost almost nothing, and are easy to store. Not relevant here, but for soldering TSSOPs, MSOPs, etc., take a look at a YouTube instructional video.

Block Diagram

Block diagram of the PA3AJR board

Technical Description

Below, the circuit will be examined, highlighting design details. A high-resolution PDF of the schematic is available in the attachment to this article. In the technical description below, fragments will be highlighted.

LNA, LNA Bias, ESD- / Overload Protection

LNA and preselection

The RF section of the receiver board consists of an SAV-514+ LNA, the bias circuit for the LNA, and ESD- / overload protection.

ESD- and Overload Protection

Directly at the input, two BAS70 diode pairs provide ESD protection without significant degradation of the noise figure. Schottky diodes were chosen for their 'soft-clipping' characteristics and mild impact on intermodulation. At the output of the LNA, three more diodes are included. These primarily serve to protect the downstream mixer and crystal filter from the relatively high RF voltages that can occur due to the high linearity of the SAV-541+.

The SAV-541+ LNA

Initially, I considered an LNA based on an Avago (now Broadcom) GaAs HEMT, the ATF-53189, but had reservations. This was mainly due to the critical input network and the (high Q / expensive) trimmer that is a constant in designs. Providing bias voltage to the gate is also challenging for the same reason. Finally, this FET does not have a great reputation for ESD or lightning discharges and overloading due to TX power. Fortunately, there is now an excellent alternative available in the form of the SAV-541+ from Minicircuits. It's an (Enhanced-) pHEMT with an integrated "noise-matched" 50Ω input and output network, about 30dB gain, a noise figure of less than 0.25dB (!), and +33dBm IP3. Superb specifications. Unconditionally stable and trouble-free to apply. And all of this for a price under 2 euros. The 30dB gain also means that the additional MMIC from the design of IZ4BEH can be confidently omitted. I got the idea for this LNA from a publication by Gyula HA8ET, who at some point also switched from the ATF-53189 to the SAV-541+ for VHF preamplifiers. This ultimately led to the following setup.

Measures for source degradation of the SAV-541+

To achieve stability over a wide frequency range and optimize the noise properties of the SAV-541+, 'inductive source degradation' has been applied following the design of Gyula HA8ET. As seen in the "magnifying glass" in the detailed layout view, the source of the FET is not directly grounded at its pin but is grounded through a rectangular shape at some distance through several vias to the bottom layer. This introduces a slight degree of inductance and feedback to secure the mentioned goals. The air coil at the input is not particularly critical, and construction details can be found in the schematic. The SAV-514+ is powered by its own 5V LDO voltage regulator. The voltage at the drain is approximately 4V, while the current through the drain is set to (nominal) 60mA. This results in a voltage between the test points TP1 and TP2 of 75mV, including the current through the bias circuit. Further details can be found in HA8ET's article.

The 144-148 MHz Bandpass Filter

Immediately after the LNA, there is a Chebyshev bandpass filter for 144-148MHz, consisting of three parallel circuits with top coupling and capacitive voltage dividers at the input and output. The filter is composed of three Neosid 10V 1 Coil Assemblies. By the way, the LC Filter design Tool from Marki has proven to be useful for designing the filter. Below is the result of the passband characteristic, measured from the input of the board to the ring mixer.

RF Chebyshev filter 0 - 290 MHz
RF Chebyshev filter Span: 1 MHz/div

In the left image, the -50dB points are located at 130 and 165 MHz, respectively. The right image shows the frequency range of 144-146 MHz. At the extreme boundaries, the attenuation is less than 0.5 dB. For the European band allocation, the filter is centered on 145MHz, but it is possible to choose 146MHz and set it slightly wider to cover the entire 144-148 MHz. Behind the Chebyshev filter, a 3dB attenuator is included to provide broadband termination for the input port of the mixer.

Coil placement

Above is a detailed view of the coil placement for illustration.

Mixer and Post-mixer Amp

Mixer and Post-mixer Amp

The traditional so-called "plug-in" models of Mini Circuits (such as the IE500, SBL-/SRH-/RAY-series) all have the same "A0x"-footprint and connections but mainly differ in their ability to handle strong signals and, consequently, the required LO power. After amplification in the LO-Amp - which will be discussed later - there is more than 23 dBm available. The prototype board is equipped with a Mini Circuits SBL-1. For this, 7 to 10dBm required LO power is specified. A 13dB PI attenuator ensures this. This can, of course, be adjusted for other types of mixers.

It is important that all three ports of the mixer are reasonably terminated over a wide frequency range. At the RF port, this is achieved with a 3dB attenuator. The port that is most sensitive to proper termination is the IF port. The matching network there serves several functions. Firstly, the 50Ω of the (10.7MHz) IF signal needs to be transformed with L8 and C24 to a value that is acceptable for the downstream MOSFET (about 39k of R14 in parallel with the gate capacitance of about 2pF). Additionally, the diodes in the mixer need to find a ground potential to prevent intermodulation. This DC return is provided by choke coil L7. Finally, it is essential that the VHF and UHF products that find termination in the mixing process, with the main component being the (2frf + fif) spurious frequency at around 280 MHz. The RC network R11//R12-C23 provides an impedance that gradually approaches 50Ω with increasing frequency.

L8 and L9 are standard Toko 10.7 MHz 10x10mm transformers, where only the primary winding is used. The capacitors on the bottom side are removed and replaced with external SMDs on the board, allowing for adjustments afterward. The mixer post-amp is a low-noise BF998,235 dual-gate MOSFET in SOT143B package. The gain is approximately 26dB. The AGC on gate 2 can reduce this to about -3dB at 0 Volt. Beware: there is also a BF998R, but the 'R' stands for 'reversed,' so it should be soldered upside down on the board!

Around the mixer, U.FL SMT test points are available to measure or inject signals.

Crystal Filter, AM/FM-SSB IF Amps and Noise Blanker

Crystal Filter, AM/FM-SSB IF Amps, and Noise Blanker

As mentioned earlier, consideration has been given to a range of possible 10.7 MHz crystal filters. The specifications of the (discrete) crystal filters available to me could not be determined. Based on similar models, frequency, and bandwidth, it seemed likely that I should assume an impedance of approximately 900Ω // 20pF. Monolithic models usually have a higher impedance, up to 4kΩ // 1pF. Since it was unsatisfactory to make an assumption, I searched for a method to determine it. I came across a methodology by Kerr Smith on This email address is being protected from spambots. You need JavaScript enabled to view it. with a reference to a post by Chuck WD4HXG on the eham forum. Based on that, I created the accessory below and performed some measurements.

Impedance Measurement of Crystal Filter

Crystal Filter Test Jig

IF Output Amplifiers and Noise Blanker

The output amplifiers for AM/FM and SSB/CW are almost identical. They are both based on a BF998 MOSFET, share the DC bias for gate 1, and both have a TOKO 10.7MHz transformer to transform the output impedance at the drain to 50Ω. However, only the SSB/CW amplifier is equipped with AGC on gate 2 and implemented with a BA591 for the noise blanker function. The main task of the amplifiers is to match the approximately 1kΩ impedance at the output of the crystal filter to 50Ω. The voltage gain is therefore a modest 8dB. Under the influence of AGC, this can be reduced to approximately 10dB attenuation in the case of the SSB/CW amplifier.


LO Amplifier

The VCO signal is applied to pin 17 of the board from the PLL unit. The frequency ranges from 123.3 to 127.3 MHz, and the measured level in my set is 3.5 dBm (330mVRMS). This level is insufficient to drive a diode ring mixer, which requires between 7 and 17 dBm, depending on the type. The 'level-23' mixers are excluded for this application. In the Mutek board, a BF274 was used to amplify the signal. IZ1DYE and IZ4BEH opted for BFQ18 and BFQ196, respectively, both in SOT89 packages. I chose a BFG591, a robust low-noise RF transistor in an SOT223 package. The output impedance of the transistor is transformed to 50Ω with a tuned circuit in the collector, a capacitive voltage divider, followed by an LC low-pass matching network.

Below is the amplifier and the matching network in practice:

LO Amplifier Characteristic
Matching Network to 50Ω

Second Revision of the Circuit Board

Below are some comments and suggestions based on experiences with the prototype and possible changes for a follow-up version. Whether such a follow-up will materialize also depends on interest. Reproducing the PCB's is not an issue, but it does require a minimum order quantity of five units and additional transportation costs. Below is a list of what I would change in a second run.

For various reasons, I would like to replace the Neosid coils with Molded coils in a revision of the print.

- The Neosid coils currently used were available to me, but they are now less common, relatively expensive, and difficult to obtain.
- Additionally, they require hand winding, making them less reproducible.
- Without an adhesive that sufficiently maintains the Q of the coils, I find them too susceptible to mechanical variations and temperature changes.

All these disadvantages would be overcome by switching to a ready-made 'molded' coil. My attention has been drawn to look-a-likes of the Coilcraft Uni-7 series. I have since ordered several and replaced the Neosid coils on the prototype with them. I am not disappointed.

They were found to be supplied with a core of both ferrite and brass, allowing the inductance to be adjustable over a wide range (measured: 54-83nH with brass core and 84-240nH with ferrite core). Also useful for other applications. They fit neatly into a 10x10mm housing from discarded TOKO transformers. All in all, they are much easier to process, more consistent, and significantly more cost-effective.

An additional footprint for mixers in an SMD package

In a follow-up, I would also like to take the opportunity to be able to use SMD-variants of the mixer in "Case Style:CD636" such as the Miniciruits ADE-series. However, it is not possible - as with the crystal filter - to project different footprints "on top of each other." Therefore, they are not 'merged' but 'placed next to each other'. In the schematic, they both appear, but of course, only one of the versions is "populated." By the way, the prices of the ring mixers are a fraction of what they were in the early '80s.

A few smaller changes will also be implemented:
  • For the top coupling in the RF bandpass filter, the single capacitor is replaced by two in series, enabling the use of standard values below 1pF.
  • The test points around the mixer in the current design are implemented with miniature (U.FL) coaxial connectors. In practice, they prove to be too cumbersome and fragile. In the print revision, through-hole test loops will be used as they are more user-friendly.
  • The 78M05 LDO is currently powered from 13.5V and dissipates approximately (13.5 - 5) * 70mA = 595mW, which can cause the temperature in the compartment to rise, potentially resulting in tuning drift. However, we still have 8V RX available on pin 11 of the board. With this, the dissipation reduces to (8 - 5) * 70mA = 210mW, which is much more favorable. In the revised board, the LDO has been moved and a jumper has been added that allows you to choose between 8V or 13.5V input voltage.
  • Given that the BFG591 in the LO amplifier is designated as obsolete by NXP and is no longer available through official distribution channels, I assume that the units sold on AliExpress are counterfeit. However, they faithfully adhere to the simulation models and perform adequately in practice. It may be considered to replace the transistor with a BFU590Q. Physically, it is a drop-in replacement, but the specifications do vary somewhat. SPICE simulations also revealed that a slight impedance adjustment at the input yields higher gain, and the pi-filter in the output, along with all its adjustments, can be omitted.
  • The trimpot between LNA and bandpass filter can be eliminated. The remaining series resistance in the layout will be replaced by a PI attenuator, but populated as a simple through-connection (0dB).
  • All pins that do not carry RF signals (like 13.5V, 8V, AGC, and NB) are proactively (and optionally) equipped with an RFI filter consisting of a (603) ferrite chip bead and decoupling capacitor.

Below is a 3D model of the revised version of the board in which all these changes have been implemented, shown with an SMD mixer, a subminiature crystal filter.


Only a minor modification is needed to make the board work in the FT221/225. This involves connecting pin 3 to the 13.5V supply voltage on receive only. The original board operates at 8V, which imposes limitations on the dynamic range. A detailed description of the procedure can be found in the MuTek manual or that of IZ1DYE or IZ4BEH attached to this article. After installation, it is necessary to readjust the zero and maximum deflection of the S-meter. As mentioned earlier, it is also worthwhile to perform the PLL unit modification as described by SM5BSZ.