Starling-V (the Final Cut)

Seas ER18RNX with 27TDFC

The first Starling I worked on was in 2010. In total, I did three versions with these drivers, the Seas ER18RNX and the 27TDFC. I was never truly satisfied with all of them. There were multiple issues that I couldn’t resolved entirely. One of which is the Seas tweeter, the 27TDFC. This led me to change to the Peerless H26TG45-06 (Starling-TPC) and subsequently, the Morel MDT29 (Starling-2020). While replacing the tweeter did a better job, I was still left wanting with the 27TDFC.

I am pleased to say that in the Starling-V, I finally sorted out the ER18RNX and the 27TDFC. This combination is incredibly difficult. The ER18 on her own is bad enough. Throw in the 27TDFC, it’s even worse. But it can be done. Here’s how I did it.

Fig 1 – Seas ER18RNX RAW Response • Baffle Width=8.75″

Fig 1 is the RAW response of the ER18RNX. This is one scary plot. She takes off like a rocket at 400Hz. Her best feature is her midrange. She’s relatively flat from 500Hz to 3kHz. After that, she breaks up with a peak at 5kHz.

There are two challenges I faced when working with the ER18RNX. The break-up peak is one of them. The other is the tonality. Her midrange is very “hot”. It needs to be lowered to match the bass. This is where it gets difficult. Lower the midrange too much and the Starling will sound terrible. Not enough, the bass is too soft. I’m talking about 1dB making or breaking the speaker.

Fig 2 – Seas ER18RNX with Low Pass Filter (LPF)

After a few attempts, I found the right combination. The Blue plot in Fig 2 is my final response of the ER18RNX. The measurement below 500Hz includes reflections in my room. Disregard the suck out at 150Hz. That’s caused by a floor bounce.

This plot allows me to compare the midrange with the bass. The midrange is about 2dB higher than the bass (120Hz to 60Hz). This is critical to how the Starling will eventually sound like. This measurement was made with the Starling-V in Full Space (4 pi).

Fig 3 – Seas 27TDFC Frequency Response

Up next is the Seas 27TDFC. Seas plot is smoother than mine. I’m getting two bumps, one at 1.8kHz and the other at 3.5kHz. They look like diffractions but my tweeter is flush mounted. That can only mean it’s from the baffle edge. I will live with this for the time being. It may not be that offensive when the Starling-V is finished.

Fig 4 – Seas ER18RNX Low Pass with 27TDFC High Pass

Fig 4 shows the ER18RNX and the 27TDFC crossing at 2kHz. I needed to cross this low because I wanted to avoid using a notch filter for the peak at 5kHz. 

Fig 5 – Starling-V Crossover Passband

The Black plot in Fig 5 is the summation of the ER18RNX and the 27TDFC. Note the slight cancellation from 5.5kHz to 7kHz. That comes from the 5kHz peaks in the ER18RNX. I can have a flatter response if I EQ the ER18 peaks away but that would increase the cost of the crossover. Since it’s quite high up in the treble, I don’t think it’s that serious.

Fig 6 – Starling-V Frequency Response • Below 500Hz in Nearfield

The Starling-V frequency response is in Fig 6. She is incredibly flat except for two blemishes. There’s a light bump between 3kHz to 4kHz which originates from the 27TDFC. After that is the cancellation seen at 5.5kHz to 7kHz. 

Fig 7 – Starling-V Null

The Starling-V Null is not the prettiest but it’ll have to do. There’s no point having a technically perfect null but losing the sound.

Fig 8 – Starling-V Step Response

The ER18 is fast for a 6-1/2″ midwoofer. There is an insignificant glitch at -25% but after that, she’s linear all the way, hitting the apex at at 250 microsec.

Fig 9 – Starling-V Waterfall

The Waterfall plot in Fig 9 shows artifacts at 5.5kHz. This is likely caused by the ER18 cone breakup peak. Beyond 5.5kHz, the 27TDFC is incredibly clean.

Fig 10 – Starling-V Toneburst Energy Storage

The Toneburst plot shows the excess energy at 5kHz last for only 8 cycles. They will not cause any trouble in the treble. It is at 1kHz that caught my attention. No stored energy is recorded there. 

Fig 11 – Starling-V Spectrogram

I was right not to use a notch filter for the ER18. The stored energy at 5kHz is fully dissipated by 1 msec (Fig 11). Two streaks are recorded, one at 1kHz and the other at 1.3kHz. They are -50dB less than the fundamental and moreover, they vanished by 6 msec. We’re not going to hear anything from them.

In Summary

The Starling-V is the best version of the ER18RNX with the 27TDFC. It is very close to the Myna that uses the U18RNX. Where the Myna beats the Starling is in the bass. I can’t see how I can improve the Starling-V further except to use a waveguide for the 27TDFC. But as it stands now, I’m at least at peace. Took me 10 years to get her to sing.

Crossover is available on request. Free for DIY. Not for Commercial use.
Unless otherwise stated, all measurements were made with the mic at 36 ins, tweeter axis. Impulse Window=5ms. No smoothing applied.