- Joined
- Jan 16, 2026
- Messages
- 12
- Likes
- 59
Introduction
S-90 is a loudspeaker model that was very popular in Eastern Europe during the 1980s. Over a million (!) pieces were manufactured by the Radiotehnika factory in Riga, Latvia, between 1980 and 1987. Many people still use them in their audio systems, argue about sound quality, and offer various upgrades.
The context, which is important to understand, is that any consumer electronics manufacturing in the command economy faced many standards and restrictions. Rationalization, cost savings, and reliability requirements in mass production also led to compromises. So, despite the factory having a world-class acoustic engineering school and its own huge anechoic chamber, the final product may be very reliable, but far from what the team could achieve acoustically.
The issue with most upgrades offered by enthusiasts is that they lack objective data to identify the real problems. Despite thousands of forum threads, the best source of measurements available online until now was the Hungarian Hi-Fi magazine from the year 1985. Several frequency characteristics of original speakers, but unfortunately, in low resolution, are possible to find on this archive page.
The goal of this article is to present Spinorama measurements of a loudspeaker with original drivers and components, analyse issues, offer listening tips, and provide a 10-band parametric equalizer to optimize the sound. These recommendations may not be a perfect solution for everyone due to mediocre repeatability over the years of production and the product lifespan. However, as we will see, the overall pattern of problems is close enough to the ones seen in the Hungarian review.
Frequency Response
To achieve qualitative results, the measurement conditions were as follows:
As we see, this option does not compensate for the baffle step and creates excessive brightness. An obvious peak in the 1-2 kHz region makes the midrange sound too forward. Then a sharp dip around 3 kHz, and a rising tweeter up to 12 kHz, where the loudest peak appears. That is the same situation as we see on the Hungarian Hi-Fi magazine review graphs.
The preference rating (assuming method restrictions for low-frequency resolution) is 0.8. By using the switch and attenuating both the midrange and high frequencies by -3 dB, it is possible to achieve a preference rating of approximately 3.1. The graph below illustrates this option:
The last option measured fully is the -6 dB attenuator position for both drivers, with a preference rating of -0.2:
For all other options, the simulation uses a phase-linear filter merge of two different measurements, and the preference rating is as follows (midrange/tweeter):
Directivity
The directivity is not bad overall, thanks to off-center driver placement, but it has some issues. The crossover frequency between midrange and tweeter is too high (5 kHz); as a result, the midrange driver's beaming causes a directivity dip at 4 kHz. Another problem is the tweeter, where the phase plug causes a diffraction effect on the driver: a directivity dip followed by a sharp rise.
To slightly reduce the effect of those issues, the recommendation is not to listen to the speakers perfectly on-axis. A small angle usually compensates for beaming and diffraction problems, which is even more important for the following equalized version.
Bass extension
Compared to most vintage speakers, the S-90 has a large cabinet and a 10-inch driver with a low resonant frequency of approximately 30 Hz; as a result, the low-frequency extension looks quite acceptable. The estimated graph based on indoor ground-plane measurements:
To achieve the best possible compromise between on-axis and off-axis response, equalization of the predicted in-room response curve is required. In that case, the overall absolute deviation from the trend line and the recommended zones will be quite small for both the on-axis and sound power curves (a small increase on-axis and a small dip in sound power, and vice versa).
The base for correction is a standard 0-0 dB switch position. The graphs below demonstrate equalization results on spinorama curves and the estimated in-room response curve with the target line of 0.76 dB/oct:
The preference rating of this configuration is 7.2 on this exact speaker sample using the following equalizer:
Conclusions
Radiotehnika S-90 speakers received recognition in the 1980s for their reliability and affordable price. And it is still relevant these days. The sound quality, however, has obvious issues. Among these issues, the most sensitive are the non-linear frequency response, midrange driver beaming, and high-frequency diffraction. To properly target those issues, the qualitative measurement is essential.
By implementing a compromise between on-axis and off-axis responses, it is possible to achieve a very smooth in-room response in most typical conditions. The parametric 10-band equalizer can help to implement such a correction without drivers or crossover replacement. Thanks to this simple solution, users can upgrade the speaker's sound quality to a level comparable to most modern systems.
S-90 is a loudspeaker model that was very popular in Eastern Europe during the 1980s. Over a million (!) pieces were manufactured by the Radiotehnika factory in Riga, Latvia, between 1980 and 1987. Many people still use them in their audio systems, argue about sound quality, and offer various upgrades.
The context, which is important to understand, is that any consumer electronics manufacturing in the command economy faced many standards and restrictions. Rationalization, cost savings, and reliability requirements in mass production also led to compromises. So, despite the factory having a world-class acoustic engineering school and its own huge anechoic chamber, the final product may be very reliable, but far from what the team could achieve acoustically.
The issue with most upgrades offered by enthusiasts is that they lack objective data to identify the real problems. Despite thousands of forum threads, the best source of measurements available online until now was the Hungarian Hi-Fi magazine from the year 1985. Several frequency characteristics of original speakers, but unfortunately, in low resolution, are possible to find on this archive page.
The goal of this article is to present Spinorama measurements of a loudspeaker with original drivers and components, analyse issues, offer listening tips, and provide a 10-band parametric equalizer to optimize the sound. These recommendations may not be a perfect solution for everyone due to mediocre repeatability over the years of production and the product lifespan. However, as we will see, the overall pattern of problems is close enough to the ones seen in the Hungarian review.
Frequency Response
To achieve qualitative results, the measurement conditions were as follows:
- Gated measurements with resolution down to 200 Hz
- CLIO microphone with an individual calibration file
- The reference axis is the center of the tweeter
- 2m distance from the loudspeaker
- Sensitivity adjusted to 2.83V/1m
As we see, this option does not compensate for the baffle step and creates excessive brightness. An obvious peak in the 1-2 kHz region makes the midrange sound too forward. Then a sharp dip around 3 kHz, and a rising tweeter up to 12 kHz, where the loudest peak appears. That is the same situation as we see on the Hungarian Hi-Fi magazine review graphs.
The preference rating (assuming method restrictions for low-frequency resolution) is 0.8. By using the switch and attenuating both the midrange and high frequencies by -3 dB, it is possible to achieve a preference rating of approximately 3.1. The graph below illustrates this option:
The last option measured fully is the -6 dB attenuator position for both drivers, with a preference rating of -0.2:
For all other options, the simulation uses a phase-linear filter merge of two different measurements, and the preference rating is as follows (midrange/tweeter):
- 0/-3 dB: 2.7
- 0/-6 dB: 0.6
- -3/0 dB: 3.0
- -3/-6 dB: 0.3
- -6/0 dB: 2.8
- -6/-3 dB: 2.4
Directivity
The directivity is not bad overall, thanks to off-center driver placement, but it has some issues. The crossover frequency between midrange and tweeter is too high (5 kHz); as a result, the midrange driver's beaming causes a directivity dip at 4 kHz. Another problem is the tweeter, where the phase plug causes a diffraction effect on the driver: a directivity dip followed by a sharp rise.
To slightly reduce the effect of those issues, the recommendation is not to listen to the speakers perfectly on-axis. A small angle usually compensates for beaming and diffraction problems, which is even more important for the following equalized version.
Bass extension
Compared to most vintage speakers, the S-90 has a large cabinet and a 10-inch driver with a low resonant frequency of approximately 30 Hz; as a result, the low-frequency extension looks quite acceptable. The estimated graph based on indoor ground-plane measurements:
- f3 at 51 Hz
- f6 at 44 Hz
- f10 at 37 Hz
To achieve the best possible compromise between on-axis and off-axis response, equalization of the predicted in-room response curve is required. In that case, the overall absolute deviation from the trend line and the recommended zones will be quite small for both the on-axis and sound power curves (a small increase on-axis and a small dip in sound power, and vice versa).
The base for correction is a standard 0-0 dB switch position. The graphs below demonstrate equalization results on spinorama curves and the estimated in-room response curve with the target line of 0.76 dB/oct:
The preference rating of this configuration is 7.2 on this exact speaker sample using the following equalizer:
- 500 Hz, A -2dB, Q 2
- 750 Hz, A 3dB, Q 2
- 1000 Hz, A -6dB, Q 1
- 1900 Hz, A -3.5dB, Q 2
- 2500 Hz, A 2dB, Q 5
- 3000 Hz, A 2.5dB, Q 10
- 3900 Hz, A -1dB, Q10
- 7000 Hz, A -4.5dB, Q 1
- 12000 Hz, A -7dB, Q 5
- 12300 Hz, A 2dB, Q10
Conclusions
Radiotehnika S-90 speakers received recognition in the 1980s for their reliability and affordable price. And it is still relevant these days. The sound quality, however, has obvious issues. Among these issues, the most sensitive are the non-linear frequency response, midrange driver beaming, and high-frequency diffraction. To properly target those issues, the qualitative measurement is essential.
By implementing a compromise between on-axis and off-axis responses, it is possible to achieve a very smooth in-room response in most typical conditions. The parametric 10-band equalizer can help to implement such a correction without drivers or crossover replacement. Thanks to this simple solution, users can upgrade the speaker's sound quality to a level comparable to most modern systems.
