Speakers Vega 50ac 106 technical characteristics of the device. Soviet acoustics. We cut, saw, wind, solder, paint and so on

As a child, one of my friends had Vega 50AC-106 speakers at home, then they sank into my soul with their design, of course, because then, in the early 90s, for me, boys, these speakers with their plastic linings were something cosmic, top perfection. Well, we didn’t think much about the quality of music then. I would like to have these for myself. Time has passed, but those speakers remain in my memory, as does the desire to receive them. Opportunities for their acquisition also appeared.

A small digression. This post does not pretend to be the ultimate truth, nor does it pretend to be any audiophile delights, nor does it claim that the result obtained breaks all records. Please don’t argue about “why are you bothering with them, it would be better to buy new normal ones”, I wanted exactly THOSE speakers, as a memory from childhood. To listen to music, I have Hifiman HE-400i headphones and a homemade tube amplifier for them; this set already claims to have some quality, although I’m not going to argue with audiophiles.

Anyone who is interested, I ask under cat.

Purchase and initial inspection

Everything turned out very simple with the purchase, I found that same friend from childhood and it turned out that he still had those same speakers. Although they are in the closet and have not been used for ten years. For a symbolic sum, we shook hands and the speakers went to me. Immediately upon purchase, I was told that one tweeter had burned out and required replacement. Plus, the speakers were pretty dusty, and there was rust on the metal grilles. Next, the speakers were connected to an amplifier to check their functionality. All speakers turned out to be working except one HF, which was confirmed by its continuity tester later. Of course, the sound of the speakers as is did not please me. After the first turn on, the speakers were completely disassembled. The case has been checked for the strength of the seams; often, with old Soviet speakers, the side panels simply come apart. I was lucky, all the joints were glued tightly; by the way, my cases were made entirely of plywood, including the back wall, which is not often found in this model. Much more often the side walls are made of plywood and the back is made of chipboard.

Next, the filter was completely disassembled to check its compliance with the original circuit, by the way, here it is:

The capacitors and resistors matched the values of the circuit. All my capacitors were metal-paper MBGO, which in general is not so bad, because in real designs people have also encountered electrolytes. By the way, after measurement, the capacitors showed deviations from the nominal value within 1%, which is simply wonderful. But most of all I was interested in the inductance of the coils, but as expected, I had no luck with them. NONE of the coils matched the circuit in terms of inductance, and moreover, the coils also did not match each other in pairs of columns. For example, the inductances of the LF coils were 0.85 mH and 0.64 mH, although according to the circuit it should be 2.8 mH. With midrange and high frequency the story is exactly the same. Well, oh well, the scheme will still be completely redone.

Work plan

After the initial inspection, a rough plan was drawn up:

1. Sealing all seams, joints and holes from the inside of the speakers with sealant;
2. Gluing wooden spacers inside the body according to the front-back and left-right pattern;
3. Treatment of all walls with rubber-bitumen mastic to reduce resonances;
4. Pasting all walls with a layer of batting;
5. Replacement of high-frequency speakers with the production of new podiums for them;
6. Complete rework of filters with changes in the circuit and part ratings;
7. Replacement of all seals under the speakers;
8. Replacement of all wires;
9. Installation of ports with terminals;
10. Washing panels and painting grilles;
11. Reduced resonances and chatter of front panels;

Well, once the plan is drawn up, you can get to work.

We cut, saw, wind, solder, paint and so on

The first step was to disassemble and clean the front panels. The metal mesh is sanded, treated with a rust converter and painted. After this, the panels were assembled back, and the metal mesh was installed in plastic with sealant to prevent them from rattling. The sealant is not visible from the outside.

Next, all the body joints and holes were carefully coated with construction silicone sealant from the inside. This is done in order to eliminate any whistling from the escaping air. There weren’t really any gaps in the body, but still. The next step was to glue two spacers from dry pickets with a cross-section of 45x20 mm into the body; they were glued between the front-rear walls and between the left-right walls, and the spacers were also glued together, so that a rigid cross was obtained. In this way we increase the rigidity of the case, because the panels here are already quite large and the wall thickness is only 12 mm. After this, the entire inside of the body is coated with two layers of rubber-bitumen mastic.

The next step was to insert ports with screw terminals for inserting wires into the housing. (for some reason the photo was not saved)

To upholster the inside of the body, ordinary batting was purchased from a fabric store, where it was about 6-7 mm thick, so mats were made from batting folded in three layers. These mats were glued to all walls except the front one, using the same rubber-bitumen mastic and additionally secured in the corners with a furniture stapler.

After finishing the work with the body, it's time to work on the electrical part. Firstly, it was necessary to decide what to install to replace the original high-frequency speakers. I was not able to find a native HF, but on the Internet they are asking quite a lot for them, from 700 to 1000 rubles for an old HF, I think it’s a lot. In addition, there is a widespread opinion on the Internet that the native 10GDV-2 in general is not particularly good. After reading the Internet and estimating the budget, as well as the feasibility of purchasing expensive tweeters for these speakers, we chose our 15GDV92-16, produced by Novosibirsk NOEMA. In terms of resistance and sensitivity, they are the same as 10GDV-2, and in terms of quality, they say, they are significantly superior to them. By the way, thanks to NOEMA for the prompt delivery and good packaging of the ordered speakers.

Looking ahead a little, I will say that I liked the sound of the speakers with these filters.
As you can see, coils with a fairly high inductance are used here at LF and MF. Winding these on “air” is not so easy, because at the same time it must also have low resistance, which means you need to use a thick wire, the dimensions of the coils are not small. Therefore, it was decided to wind the coils on frames with cores made of transformer iron. A plumbing polypropylene pipe with a diameter of 40 mm was taken as a frame; strips of transformer iron, cut from old Soviet transformers, were tightly packed inside. The bass coil is wound with a wire with a diameter of 1.6 mm over varnish, and the midrange coil is wound with a diameter of 1.2 mm over varnish. The HF coils were wound on original frames without cores, with original wire, where it is somewhere around 0.6-0.8 mm in varnish. All coils were wound with careful monitoring of the resulting inductance using an LC meter. This made it possible to achieve an almost perfect match of inductance with the circuit. The capacitors in the filters were the original MBGO ones, all resistors were ordered new with a power of 35 W, a resistance of 10 Ohms and 3.3 Ohms. 10 Ohm resistors were taken with a reserve and, using a tester, values closest to those indicated in the circuit were selected. All filter wiring is made with a rigid single-core installation wire with a cross-section of 2.5 mm2. This section was chosen not so much from the desire to make the wire as thick as possible, but so that the wire would not dangle and reliable contact pads could be bent from it.

The filters were installed in the housing, of course they did not fit in the old place, we had to re-drill the holes and lower the filter lower. The top of the filter was covered with batting; it was specially made longer on the back wall.

The bass reflex tubes had to be glued together before installation, as they fell apart into two halves in our hands. They were inserted into the body using sealant and covered with batting.

It came to the speakers. The HF are fixed to the podiums through gaskets. The midrange cups are inserted into the housing using sealant. The glasses themselves are half filled with fluffy cotton wool. The midrange and bass speakers are covered with seals made from the automotive noise-insulating material Splen. When pulling the speaker, it compresses the joint very well. When pressed, the midrange speaker goes inward reluctantly and returns slowly, which indicates the tightness of the structure.

Unfortunately, time has not spared the shiny metal caps on the woofer and midrange speakers; they have peeled off slightly. I had to paint them with acrylic spray paint. I tried to do this as carefully as possible and in as thin a layer as possible.

And to complete the work on the body, the legs were cut out to replace the poor plastic ones. I cut legs with a diameter of 7 cm from a sheet of soft rubber 4 cm thick. And let audiophiles throw rotten eggs at me, but they work no worse than spikes, which I abandoned after much thought.

After assembly, this is what happened:

After that, the entire body was covered with colorless matte varnish.

The inside of the plastic facing panels were covered with pieces of STP Aero automotive vibration insulation, this gave them weight and completely removed all the ringing. Madeline strips were glued under the panels to the body; this is a special sealing anti-creaking material, something like impregnated foam rubber. This was done in order to press the panels as tightly as possible to the body.

The set results have been achieved. I became nostalgic for those same speakers from my childhood, achieved better and more pleasant sound and had a pleasant time with my hobby. Now these speakers are in my home as the main ones, but it is far from a fact that after a while they will not be replaced by something of better quality, and it is quite likely that these could be other old Soviet speakers.

The budget for the entire event did not exceed 5,000 rubles, this included the purchase of speakers, the purchase of speakers and the purchase of some materials. For this money, I think it’s impossible to find something of better quality and better sound. By the way, for the same S90 Radio Equipment they ask us for 10,000 :) They promoted them on the Internet :)

Thank you all for your attention!

You can help and transfer some funds for the development of the site

Introduction

Photo not mine, taken

Purchase and initial inspection
Everything turned out very simple with the purchase, I found that same friend from childhood and it turned out that he still had those same speakers. Although they are in the closet and have not been used for ten years. For a symbolic sum, we shook hands and the speakers went to me. Immediately upon purchase, I was told that one tweeter had burned out and required replacement. Plus, the speakers were pretty dusty, and there was rust on the metal grilles. Next, the speakers were connected to an amplifier to check their functionality. All speakers turned out to be working except one HF, which was confirmed by its continuity tester later. Of course, the sound of the speakers as is did not please me. After the first turn on, the speakers were completely disassembled. The case has been checked for the strength of the seams; often, with old Soviet speakers, the side panels simply come apart. I was lucky, all the joints were glued tightly; by the way, my cases were made entirely of plywood, including the back wall, which is not often found in this model. Much more often the side walls are made of plywood and the back is made of chipboard.
Next, the filter was completely disassembled to check its compliance with the original circuit, by the way, here it is:

The capacitors and resistors matched the values of the circuit. All my capacitors were metal-paper MBGO, which in general is not so bad, because in real designs people have also encountered electrolytes. By the way, after measurement, the capacitors showed deviations from the nominal value within 1%, which is simply wonderful. But most of all I was interested in the inductance of the coils, but as expected, I had no luck with them. NONE of the coils matched the circuit in terms of inductance, and moreover, the coils also did not match each other in pairs of columns. For example, the inductances of the LF coils were 0.85 mH and 0.64 mH, although according to the circuit it should be 2.8 mH. With midrange and high frequency the story is exactly the same. Well, oh well, the scheme will still be completely redone.

Work plan
After the initial inspection, a rough plan was drawn up:
1. Sealing all seams, joints and holes from the inside of the speakers with sealant;
2. Gluing wooden spacers inside the body according to the front-back and left-right pattern;
3. Treatment of all walls with rubber-bitumen mastic to reduce resonances;
4. Pasting all walls with a layer of batting;
5. Replacement of high-frequency speakers with the production of new podiums for them;
6. Complete rework of filters with changes in the circuit and part ratings;
7. Replacement of all seals under the speakers;
8. Replacement of all wires;
9. Installation of ports with terminals;
10. Washing panels and painting grilles;
11. Reduced resonances and chatter of front panels;

Well, once the plan is drawn up, you can get to work.

We cut, saw, wind, solder, paint and so on
The first step was to disassemble and clean the front panels. The metal mesh is sanded, treated with a rust converter and painted. After this, the panels were assembled back, and the metal mesh was installed in plastic with sealant to prevent them from rattling. The sealant is not visible from the outside.

The next step was to insert ports with screw terminals for inserting wires into the housing. (for some reason the photo was not saved)

While the speakers were traveling through Russian Post, I started reworking the filters. Again, at first I dug through the forums for a very long time, collected opinions, and eventually found a scheme for altering the filters of these speakers, with good reviews and sufficient justification from the author.
The scheme was taken

Looking ahead a little, I’ll say that I liked the sound of the speakers with these filters.
As you can see, coils with a fairly high inductance are used here at LF and MF. Winding these on “air” is not so easy, because at the same time it must also have low resistance, which means you need to use a thick wire, the dimensions of the coils are not small. Therefore, it was decided to wind the coils on frames with cores made of transformer iron. A plumbing polypropylene pipe with a diameter of 40 mm was taken as a frame; strips of transformer iron, cut from old Soviet transformers, were tightly packed inside. The bass coil is wound with a wire with a diameter of 1.6 mm over varnish, and the midrange coil is wound with a diameter of 1.2 mm over varnish. The HF coils were wound on original frames without cores, with original wire, where it is somewhere around 0.6-0.8 mm in varnish. All coils were wound with careful monitoring of the resulting inductance using an LC meter. This made it possible to achieve an almost perfect match of inductance with the circuit. The capacitors in the filters were the original MBGO ones, all resistors were ordered new with a power of 35 W, a resistance of 10 Ohms and 3.3 Ohms. 10 Ohm resistors were taken with a reserve and, using a tester, values closest to those indicated in the circuit were selected. All filter wiring is made with a rigid single-core installation wire with a cross-section of 2.5 mm2. This section was chosen not so much from the desire to make the wire as thick as possible, but so that the wire would not dangle and reliable contact pads could be bent from it.

While I was fiddling with the filters, which took more than one day, the speakers arrived. Of course, their dimensions were different from the old ones, so we had to make new podiums for them. Well, after that it came to assembly.
The filters were installed in the housing, of course they did not fit in the old place, we had to re-drill the holes and lower the filter lower. The top of the filter was covered with batting; it was specially made longer on the back wall.

The bass reflex tubes had to be glued together before installation, as they fell apart into two halves in our hands. They were inserted into the body using sealant and covered with batting.
It came to the speakers. The HF are fixed to the podiums through gaskets. The midrange cups are inserted into the housing using sealant. The glasses themselves are half filled with fluffy cotton wool. The midrange and bass speakers are covered with seals made from the automotive noise-insulating material Splen. When pulling the speaker, it compresses the joint very well. When pressed, the midrange speaker goes inward reluctantly and returns slowly, which indicates the tightness of the structure.

Unfortunately, time has not spared the shiny metal caps on the woofer and midrange speakers; they have peeled off slightly. I had to paint them with acrylic spray paint. I tried to do this as carefully as possible and in as thin a layer as possible.
And to complete the work on the body, the legs were cut out to replace the poor plastic ones. I cut legs with a diameter of 7 cm from a sheet of soft rubber 4 cm thick. And let audiophiles throw rotten eggs at me, but they work no worse than spikes, which I abandoned after much thought.
After assembly, this is what happened:

After that, the entire body was covered with colorless matte varnish.
The inside of the plastic facing panels were covered with pieces of STP Aero automotive vibration insulation, this gave them weight and completely removed all the ringing. Madeline strips were glued under the panels to the body; this is a special sealing anti-creaking material, something like impregnated foam rubber. This was done in order to press the panels as tightly as possible to the body.

At this point the work was completed. Based on the listening results, we can say that the speakers began to play much better than before the modification, they became a pleasure to listen to. There will be no comparisons simply because I have nothing to compare with. There are Microlab SOLO2 Mk2 speakers converted into passives. The same tracks from the same amplifier are played much better by the converted Vega, but this is not surprising, the class of these microlabs is not at all high. Comparing Vega with Hifiman HE-400i headphones is at least stupid, these are things of completely different classes and different eras.
The set results have been achieved. I became nostalgic for those same speakers from my childhood, achieved better and more pleasant sound and had a pleasant time with my hobby. Now these speakers are in my home as the main ones, but it is far from a fact that after a while they will not be replaced by something of better quality, and it is quite likely that these could be other old Soviet speakers.

Anyone who is interested, I ask under cat.

Purchase and initial inspection

Work plan

After the initial inspection, a rough plan was drawn up:

Well, once the plan is drawn up, you can get to work.

We cut, saw, wind, solder, paint and so on

The next step was to insert ports with screw terminals for inserting wires into the housing. (for some reason the photo was not saved)

The bass reflex tubes had to be glued together before installation, as they fell apart into two halves in our hands. They were inserted into the body using sealant and covered with batting.

After assembly, this is what happened:

After that, the entire body was covered with colorless matte varnish.

Thank you all for your attention!

I decided to take these speakers apart and look inside. Disassembled.
Removed filters
Manufacturing defects that had to be eliminated:
1. I coated all the seams with plasticine (this eliminates overtones and losses in the low frequencies)
2. Installed stiffening ribs on the walls of the speakers (reduces radiation from the walls of the speaker and resonances at low frequencies)
3. Covered the inner walls of the speaker with a layer of synthetic padding (additionally reduces overtones and resonances)
4. Adjusted the offset of the acoustic center of the speakers.
5. Replaced the internal wiring with a thicker wire 2 mm2
6. Changed the filter scheme
7. Replaced the tweeter with a ribbon tweeter 10 GI-1
To adjust the displacement of the acoustic center of the woofer, we make a stand. Cut from plywood. You can also use several layers of fiberboard for this.
It is glued with PVA glue at the location where the woofer is installed and shifts its position 1.5 cm outward. As a result, the acoustic centers of the high-frequency, mid-range and low-frequency speakers are at the same level.

This is what the speaker offset looks like in a side view

I installed modern terminals, these are the ones I like best, and you can clamp the wire and fit the plugs.

Here is a view of the converted speakers without the decorative nameplate.

These are the original filter circuits that had to be upgraded to get decent sound.

The result is this.

Since changes in the circuit were made by ear, therefore, what harmed the sound was removed and what improved it was added. A piece of 2000NM ferrite ring is inserted into the low-pass filter coil (to increase the inductance), but it is better to fill it with permalloy - take plates from a network transformer. Such tricks were made to get by with minimal alterations and the use of components available in the filters. The resistance is in series with the midrange speaker of a fairly large value - this is to balance the increased output on the midrange. All parts are used from original filters. The only thing I had to add was powerful resistances when selecting 25 Ohms.
Traditional low-pass and mid-range filters with a capacitor installed parallel to the speaker make the sound rather faded, I think this is the influence of the capacitor connected in parallel to the speaker, which, forming a resonant circuit with the speaker coil, worsens the microdynamics of the sound. But this is purely my opinion, so listen and check for yourself, perhaps you will like the sound from traditional filters more.

1. History and overview of existing modification options

2.1. Using RMAA

3.1. Multi Meter 0.03

4.1. Frame
4.2. LF section
4.3. Midrange section
4.4. HF section
5. Plans

Part 1
1. History and overview of existing modification options
Based on the AS "Vega 50AS-106"

35GDN-1-8, or, according to the old GOST, 25GD-26, which began to be produced in the early 70s as a development of the 10GD-30 - this was the first domestic compression-type head, size 20 cm (8"). Characterized by a rather low resonance frequency (30…40 Hz), low impact (0.12…0.15 Pa, or, according to modern standards, 85-86 dB/W/m). The first series 25GD-26 had a magnet with rare earth metals Version “B”. had a regular magnet. There were both 4 and 8 Ohm versions.
Based on the 25GD-26, quite a few three-way speakers were created, 25AS-2, 25AS-11, 25AS-111, 25AS-109, 25AS-126, later, with new names - 50AS-103, 50AS-104 and 50AS- 106 "Vega", S50B, and two-way - 15AC-404.
At first, 6GD-6 or 10GD-34 were used as a midrange link in industrial speakers, after the start of production of 15GD-11, it began to be used everywhere in several modifications, later called 20GDS-3, 4.
The HF link was 3GD-31 (5GDV-1), in later versions - 10GD-35 (10GDV-2, 6GDV-6).
Acoustic systems based on the described DG set were produced with a “closed box” design, for example, 25AS-109, or a bass reflex - this mainly refers to later developments, 25AS-109-1 (50AS-103), 25AS-109 -2 (50AC-104). The useful internal volume ranged from 12 liters (15AS-404, ZYa) to 40 (25AS-111, ZYa).
Most often, “typical” crossover frequencies of 500 and 5000 Hz are promised. Less often, 200 and 5000 Hz - but this is already in three-component systems, according to modern 2.1, when the low-frequency heads are located in a separate box in the form of a bedside table, and the midrange and high-frequency are in the form of bookshelf speakers.
Over the course of more than two decades, quite a few options for redesigning and fine-tuning this family of acoustic systems have been published. First of all, it is recommended to eliminate factory defects: ensure the tightness of the case, perform vibration and sound absorption, replace the supply and internal cables.
Further modifications can be done at little additional cost. It is recommended to modify the filters to obtain more accurate crossover frequencies.
Speakers with a closed design and a small volume are recommended to be supplemented with a bass reflex to obtain greater output from the low-frequency section.
It is recommended to modify the midrange and treble heads.
The 15GD-11 used has a low threshold sensitivity due to the low flexibility of the moving system and the heavy diffuser, depriving the phonogram of the quietest sounds and, in addition, produces a lot of overtones, especially without the use of PAS. It is recommended to add an ARS (acoustic impedance panel) to the midrange section to suppress surges in the output of installed diesel generators at the main resonance frequency, and reduce nonlinear distortions. In addition to reducing distortion by installing a PAS, it is recommended to equip it with an additional radiating dome of sufficiently high rigidity, securing it over the “original” dust cap. They also say that gluing the cap inside out, with the convex side inward, gives almost the same result (tried it with 35GDN-1, didn’t notice).
Used for a long time as an HF link, the 3GD-31 head has an unpleasant “metallic” sound. Its modification comes down to placing a sound-absorbing coating made of felt or Guerlain inside the head and, additionally, covering the back side of the diffuser with a vibration-absorbing composition based on Guerlain or polyisobutylene (this is easy to describe, much more difficult to do...).
Thus, with a small investment of money, a fairly decent speaker sound, in comparison with the original one, is achieved.
More expensive options are to replace the dynamic heads with ones that provide better sound with appropriate adjustments or replacement of crossover filters. Often they combine the replacement of midrange and high-frequency heads with the installation of a bass reflex. Instead of the 3GD-31, you can install the more compact 4GDV-1, with minimal modifications and changes in appearance, adding a bass reflex to the design. It is recommended to use 6GD-13 (6GDV-4), as it is more sensitive and has better sound.
In the two-way version, it is recommended to use the broadband 3GD-42 instead of the 3GD-31, or convert the speaker into a three-way one.
In addition, all recommendations related to improving the 35AC-*** family with minimal adjustments (adjusting to the sensitivity of the low-frequency element) are suitable for upgrading the midrange and high-frequency sections.

I will not offer anything fundamentally new: I will simply describe my path, taking into account the experience of my predecessors and with the least financial investment.

2. Measurements. How it was for me
2.1. Using RMAA
Using this program (audio.rightmark.org), with a little desire, you can remove the frequency response of almost everything...
In relation to AS construction, you can remove the frequency response of filters, frequency response of heads, and you can also use RMAA as a voltmeter.
I won’t describe how to use the program here; that’s in the instructions.
To remove the frequency response of filters, we do everything as usual: we connect the output of the sound card to an external amplifier (if the sound card does not have a power amplifier; if it does, one will do), and the input of the sound card to the filter load.
You must remember to match the levels so as not to “kill” the sound card input and get reliable results. Therefore, it is possible that a voltage divider will be needed. However, the frequency response of the filter does not depend on the signal level, so you can make the amplifier's output signal level minimal so that the sound from the speaker is barely audible. You need to be especially careful when using an external amplifier without a volume control (actually, a power amplifier, “end”).
In versions of RMAA up to 5.1, the frequency range for the calibration tone could be set within rather narrow limits for this application, therefore, it was problematic to remove the frequency response of a high-pass filter operating even from 5 kHz, because the signal was suppressed by the filter. Starting from version 5.1, the frequency range has been expanded to 30...15000 Hz (for which, many thanks to the authors!), which is enough with a reserve. I will also note that the lower limit of 30..60 Hz can be useful when removing the IFC response of low-frequency heads, low amplifier power and a sufficiently large value of the current-setting resistor.
If you have figured out the levels, you can apply a measuring signal to the input of the filter with a load, connecting the input of the sound card to the filter load. It is advisable to immediately use a dynamic head as a load, especially if it is for a low-frequency or mid-range link: this is where the usefulness of circuits that correct the change in impedance with increasing signal frequency will become visible, and for the high-frequency link, the effectiveness of the circuit for suppressing the surge of recoil at the resonant frequency, if there is.
Scheme for measurements

Example of removed frequency response filters -

Unfortunately, RMAA does not yet provide the ability to “move” the graphs relative to each other to make it look more clear.
Removing the ICH will require a little more effort. Scheme for removing ICH

Compared to the previous one, a resistor must be connected in series with the load as a current source. The JBL Speaker Shop recommends using a 1k ohm resistor. In Vinogradova’s book, it is recommended to take this resistor with a resistance of 200...250 times the nominal resistance of the DG. On my own behalf, I will say that the denomination can be determined based on what we need to get. To determine the resonant frequency of a dynamic head or a box with a bass reflex, a resistor with a nominal value of 10 times the nominal resistance of the DG is sufficient, i.e. 75…100 Ohm. If we set a goal to measure the frequency response of the DG to calculate the Thiele-Small parameters, it is desirable that the current-setting resistor be the same 10 times greater than the maximum impedance of the DG, as a rule this is the impedance at the resonant frequency. Here it is more difficult. Of course, 1 kOhm is enough. My maximum impedance was about 60 Ohms - this depends on the quality factor of the DG and the nominal resistance. However, in practice, we may not have an amplifier with a power of 100...200 W on hand, as recommended in the description of the Speaker Shop program. Therefore, choose a compromise solution. When using an external amplifier with an output voltage of up to 15 V, I used a 1 kOhm resistor. It must be remembered that the accuracy of the results depends on the value of the current-setting resistor if measurements are needed. If the ratio of the maximum impedance of the DG and the current-setting resistor is 1:10, the measurement accuracy will be ±10%, which is sufficient, since further calculation of the parameters of the loudspeaker (AS) is rather of an estimation nature.

When using a sound card amplifier, I settled on a nominal value of 300...400 Ohms. With large resistor values, the sensitivity of the linear input of the sound card is not enough.

If you need to take a characteristic in order to obtain digital values, and not just a relative graph, you need to start from something. To do this, you will need a constant resistor with a known resistance and a second current-setting one with the same rating as the first

For ease of use of graphs, this resistance should be commensurate with the impedance of the experimental DG, i.e. lie within 4...20 Ohms. If you use a sound card with an amplifier, I recommend taking this resistor larger, 10...20 Ohms, so that the sensitivity of the linear input is sufficient. The RMAA program “looks” for a calibration signal in one channel, so the connection must be selected in such a way that the signal enters this “required” channel from this resistor, if its value is greater than the nominal impedance of the DG and for certainty. In this case, the inconvenience will be that it will not be possible to compare several graphs: when we try, we will see several straight lines. Another option is to first approximately determine the resonant frequency of the head, and select it as the frequency of the calibration tone; a fairly high frequency, above 5 kHz, is also suitable, when the impedance of the DG noticeably exceeds the value at a frequency of 1 kHz - this is where expanding the frequency range of the calibration tone comes in handy .
An example of a recorded IFC-

The green line corresponds to a 10 ohm resistor.
2.2. T-S Settings Using RMAA and JBL Speaker Shop
Now, having learned how to take the IFR of the DG, let's try to measure the T-S parameters. To do this, you will additionally need, in addition to an amplifier with an output voltage of 15...20 V, an AC voltmeter (a regular AVOmeter Ts4*** or a modern Chinese digital one will do) and another resistor with a nominal value of about 100 Ohms. Of course, an ohmmeter to measure the resistance of the DG coil.
Do not forget that digital devices costing up to approximately 1500 rubles (and maybe more, I just haven’t used more expensive ones) with closed probes at the lowest limit of resistance measurement do not show “0”, but some value, up to 0. 5 Ohms - When measuring the resistance of the DG coil, if it is around 3.5 Ohms, which is typical for heads with a nominal impedance of 4 Ohms, this can introduce a very noticeable error.

The JBL Speaker Shop program recommends using a millivoltmeter, but I don’t have one, it’s expensive, making it yourself is a hassle and it will be hard to believe, so I replaced it with the RMAA program. The trouble here is that you will need another audio frequency generator program, I used SoundForge, and a second sound card in the computer. Maybe there are programs that allow you to do both - work as a generator and a millivoltmeter, and get by with one sound card, but I don’t have one, and this decision came easily. And two sound cards, I think, are not a problem for the majority now: all modern motherboards have a built-in AC "97 codec, and most users still prefer to use something hardware with better parameters. In my computer, the motherboard has " sound" on the CM8738 chip, and an external sound card with a built-in UMZCH - Creative Sound Blaster PCI. They had to be used together, because the RMAA program uses both the input and output of the sound card when working, and cannot generate whatever your heart desires The main requirements for the characteristics of sound cards are a slight unevenness in the range of low audio frequencies. I have a combination: “output of one board - UMF - input of the second board” at a frequency of 20 Hz gave a rollover of about 3 dB, which had to be corrected even when using an external amplifier. with a sound card amplifier it will be even worse. Therefore, it is advisable to additionally have a two-position switch.

So that you can quickly make adjustments by comparing the readings on a “reference” resistor with a nominal value of 100 Ohms. That is, with a switch we switch the load of the current-setting resistor: DG or 100 Ohm resistor.

The operating procedure, in accordance with the instructions from JBL Speaker Shop, is as follows. We connect the sound card output to the amplifier, load the amplifier output with a 1 kOhm current-setting resistor, its load, in turn (DG or 100 Ohm resistor), is selected by a switch. We connect the input of the second sound card to the current-setting resistor on the load side.

Now is the first stage of calibration of our measuring system. In the RMAA device settings for “Playback/Recording”, select a sound card that is not used in the system “by default”. Sound Forge, for example, does not allow you to select the device with which it works, but RMAA does. Let me remind you that the output of the sound card in the system used “by default” is connected to the input of the amplifier, and the output of the one that RMAA uses is not connected to anything useful to us now.

We supply a signal with a frequency of 500 Hz (in Sound Forge, with the file open, Tools->Synthesis->Simple, a signal of the Sine form, lasting 20...30 seconds and click Preview. Something similar is in Wavelab, CoolEdit, etc., signal must be at the maximum level) and set the 100 Ohm resistor with the volume controls (of the sound card and, if any, of an external amplifier) to such a voltage so that, on the one hand, it does not overload the input of the sound card, on the other hand, so that it can be measured using the device we have. Inexpensive in the past, but quite expensive now, the Ts4*** devices that I came across had the smallest limits for measuring alternating voltage from 0.75 V to 2.5 V. An inexpensive Chinese pointer device would also be suitable. Do not forget that in the first third of the scale the instrument error is not standardized, therefore set the voltage so that the arrow is somewhere in the second half of the scale and coincides with some “large” division. I used the Ts4341 device with the lowest limit of 1.5 V and set the voltage to 1 or 1.5 V.

Now we look at the computer screen, select “Adjust I/O levels” in RMAA and see what we got there. If possible, use the recording level controls to select a level close to –1 dB, as the program authors advise. Make sure that the harmonic level is not too high. I didn’t have the opportunity: it was not possible for the Creative Sound Blaster PCI to receive an undistorted signal with a level greater than -4 dB, which in this case is not particularly important, and I used a reference of -6 ... -9.9 dB. Well, it is desirable that the noise level should not be higher than –40 dB compared to the level of the useful signal. In the program window you can observe not only the spectrum analyzer, but also the value of the input signal level in digital form, which we will use.

So, we adjusted the signal level at the sound card input so that it would not be overloaded and the tester would show a certain value. Let's write down the signal levels in volts and decibels. For convenience, we will use Microsoft Excel to convert decibels to volts (JBL Speaker Shop wants voltage values to be given in volts). Actually, it’s not the absolute values that are important, but their ratios, but readings in decibels still can’t be used directly, so we’ll recalculate them in Excel. Knowing one reading in volts and its correspondence in decibels, it is easy to obtain the voltage level in volts by reading what the RMAA shows (in decibels)

Difficult and troublesome? But it’s cheap: you don’t have to buy anything. If there is no tester from ten years ago, the matter becomes more complicated. There is nothing to do without it.

Well, we're done with the preparations. Now comes the characterization stage. Everything is according to the recommendations of JBL Speaker Shop (Test -> Loudspeaker), only the measured RMAA values are first substituted into the Excel table to be converted into volts.

Another little trick that you shouldn't forget about. As I already said, the “measuring equipment set” may have a large frequency response nonlinearity near 20 Hz. It can be corrected “on the fly” by sending a signal and temporarily loading the current-setting resistor with a 100 Ohm resistor, compare the readings with those at a frequency of 500 Hz and then adjust the signal level taken from the DG.

It took about half an hour to measure the parameters of one head. Theoretically, it would be possible to get by with a sound card amplifier, but then the problem of insufficient sensitivity of the voltmeter may arise, i.e. you need a millivoltmeter, and if you have one, there’s no point in worrying about RMAA.
2.3. T-S parameters "manually" using RMAA
Have you come across the article “Measuring Thiel-Small parameters at home”? (http://ussrhi-fi.ru/files/till_small.rar). And I got it. A similar thing is described in the book by Vinogradova and Voishvillo. After the ordeal with the JBL Speaker Shop, I decided to test myself with this technique. The essence is the same, only here there is a description as a hint. It takes about the same amount of time to calculate the characteristics, but this method is somewhat more visual, and you can control your every step.

Let's go back to

As mentioned above, you will need a second current-limiting resistor with a value as close as possible to the first, and a resistor with a resistance of 5...20 Ohms as a reference. An ohmmeter is needed as an external instrument. Using an ohmmeter, we find the required pair of current-setting resistors. I used a pair of 370 ohms with ±1% accuracy class. Resistors from the same batch (not defective) usually have fairly close values; if it is not possible to accurately measure the resistance, buy these.

We select the resistance of the resistors based on the same premises, choosing between the accuracy of the results and the capabilities of the amplifier. I did not use an external amplifier, but one built into the sound card. It is necessary to measure the resistance of the reference resistor as accurately as possible (if possible, use an industrial, high-precision one; if not, try to find at least an inexpensive digital device and use it to measure the reference resistor and the resistance of the DG).

First, you need to make sure that the frequency response of the “measuring complex” is linear; to do this, we connect the input and output. If large unevenness is noticed at low frequencies, write down the measurement result so that you can use it for correction later. My unevenness is noticeable, especially in the low-frequency region, so we can’t forget about it

Taking the calibration signal from the reference resistor, set “Adjust I/O levels” and carry out measurements on the “bench”. We adjust the resulting graph if necessary using the previous one.

What do we see on the graph?

One line is straight - this is from the reference resistor (possibly after adjusting the frequency response of the board, like mine), the second - with a “hump” at low frequencies and a rise from about 1 kHz. Using this graph we will carry out calculations. Microsoft Excel will help again.

Using RMAA tools, we examine individual sections of the graphs “under a magnifying glass” and transfer the data to a table.

We mark the intersection point of the two graphs simply to check the correctness of the formulas. The minimum impedance is also a good idea to know when caring about the amplifier. The maximum impedance is needed for calculations. Impedance at a frequency of 1000 Hz - for comparison with the nameplate value.
Yes, I almost forgot: you will need a school ruler to measure the diameter of the diffuser. I looked for her around the house for a long time.

Once again we repeat the procedure for obtaining a graph for a head placed in a closed volume, calculate the new T-S parameters and use them to calculate the equivalent volume. You can only get by by taking the resonant frequency of the DW in the box and using the approximate formula.

So we got a couple of graphs before the calculations. Beautiful? It's good if they are beautiful, like on

The impedance peak is sharp and symmetrical. This is how it should be. But for some reason this doesn’t always happen...

There is no such beauty... And the main thing is how to calculate from such graphs what to take for the maximum impedance? The first thing that came to mind was to finish drawing it to get a sharp peak. To do this you will need some kind of graphics editor. As it turned out later, this is what Speaker Workshop does when calculating the expected schedule. You can try to take the graphs again until you get the “correct” ones.

It took me about six months to do everything described above and try to “tuning” the Soyuz-110 stereo PA tape recorder - in the evenings and on weekends. Along the way, I read the “Cotton Branch” by iXBT.
3. Measurement of characteristics of passive elements for filters
There is one more problem. You will either need to rebuild or create new filters. What to do without devices? The first thing that came to mind: buy. However, it turned out that inexpensive ones - up to 1000 rubles - can only measure capacity. For 1500...2500 you can already buy a device that allows you to measure both capacitance and inductance with acceptable (±2%) accuracy for elements that can be used in speaker filters, but it’s a pity to spend so much money (a pair of modified speakers can be purchased for 500 rubles in a successful situation .). Then I came across the MM-Multi Meter program (http://www.i-adrian.home.ro/file/mm.zip).

We also managed to use the inexpensive M890 device, which can measure capacitance up to 20 µF. The error of the device turned out to be about ±5%, however, as promised.

Then I got my hands on the P588 AC bridge with an error of ±1%, it allows you to measure capacitance from 1 nF to 1100 μF, inductance from 1 μH to 11 H, DC resistance from 10 mOhm to 11 MOhm, in addition - AC resistance in within the same limits, but with an error of about ±5%.
Measurements are made at a frequency of approximately 1 kHz. The bridge is manual, so it takes about 5 minutes to measure one value. But it’s free... Using this device, several dozen capacitors were measured, obtained from the remains of computer equipment from the times of the USSR, mainly of the K73-1x type, in addition - MBM (1 μF each ) and MBGO, MBGCH. It is pleasant to note that K73-1x capacitors were widely used not only in the USSR, but also in production technology in the countries of the former socialist camp - therefore, the quality was trusted. The same cannot be said about electrolytes. I also measured the characteristics of the inductors that were used in the filters.

In general, the impression was that the device lives up to its stated accuracy and the measurement results can be used as a standard for testing other devices and measurement techniques.
3.1. Multi Meter 0.03
After elements with almost standard values had been accumulated, I undertook to find out the measurement accuracy of the means more accessible to me: the bridge, sooner or later, would have to be given away.
The simplest measuring tool is the Multi Meter program. It allows you to achieve quite acceptable measurement accuracy (±3%), but during operation it requires frequent replacement of the frequencies at which measurements are carried out. So it's more like fitting the result to the answer. Look at the measurement results on the “MM accuracy” page of the attached xls file.

Advantages of MM.
1. Small size of the program.
2. Wide range of capacitance measurements, from 0.01 µF to a thousand (or even more - I don’t know, I have nothing to compare with).
3. The ability to set the frequency for measuring inductance for coils with a core allows you to see its nonlinearity, if, of course, it manifests itself in the frequency range from 50 to 1000 Hz.

Disadvantages of MM.
1. Unpredictability of results. It is necessary to change the frequencies of the measuring signals for different ranges and for different elements
2. There seems to be calibration, but does it work? I didn't notice. But it would be appropriate: in order to measure the impedance of a DG with a nominal value of 4 Ohms, it would be advisable to compensate for the resistance of the wires.
3. There is no clarity with the choice of input and output levels of the sound card. I used RMAA, Adjust I/O levels, set to maximum, -1 dB or -6 dB. So you need to keep another program at hand, otherwise the predictability of the results will decrease even more.
3.2. Box and Speaker Workshop
And now I’m ready to “assemble” this famous device http://www.speakerworkshop.com/Files/SpkrworkIntl.zip. Why didn't you do it earlier? I was scared by the three-position switch, which I don’t have, and the terrible stories of experienced ones. It sounded something like “I installed Linux the day before yesterday. I almost died. Yesterday I installed Windows-95. It would have been better if I had died the day before yesterday!!!”

According to Shikhatov's description http://www.bluesmobil.com/shikhman/arts/box.htm nothing was measured. Then I came across a story from Melomana (http://dev.azz.ru/korobochka.txt) that a divider can only spoil everything. And that’s right: he ruined everything. After throwing out the divider, everything was set up and began to measure. The results are on the "Box" page. Accuracy - ±3%. True, the measurement limits are very narrow.
But, as Murphy's Law says, "Don't try to repeat a successful experiment!" (Fett's Law). The first successful experiment remains the last for now. A dozen attempts to calibrate the “box” were unsuccessful, the accuracy of the results is worse, approximately ±5-10% (in the xls file “take two”).

The box layout has been simplified a little.

One Reference resistor R1 and one calibration resistor R2 were used. I had to leave one switch to calibrate the difference in channels. By the way, if there is a divider at the input, this switch position is not needed at all (compared to the resistance of the divider resistor, as indicated by Shikhatov - 11 kOhm, a resistance of 4...10 Ohm does not make any noticeable contribution felt by our devices. So in vain I spent a long time with pulled "assembly"). In addition, it seemed to me that the connecting cables from the “box” to the measured object also make a contribution, especially when measuring low resistances. Therefore, I connect the resistor used during calibration to the same place as the object being measured. Instead of connecting a second resistor, I simply connect the probes together (for this, the Reference resistor (R1) should not be less than 4 Ohms). This does not affect the accuracy of the calibration. Either it’s fate or it’s not... Sometimes it helps to play with the levels of the input-output signal. But this is more like not adjusting the result to the answer: I can achieve the correct measurement of what I know, but how the DG parameters are measured is not clear, I don’t have a “reference” one.

I still used the Creative Sound Blaster PCI with the built-in amplifier. Attempts to use the CM8738 were unsuccessful. In principle, when recording a test signal (according to Shikhatov’s description), it was clear that in one of the channels it was greatly distorted if it was loaded onto a calibration resistor. Apparently, although the chip has a headphone preamplifier, its output current is insufficient.

I used the program version 1.06. The problems that Shikhatov writes about - about the opening of "item-a" and about the equivalent volume - have been eliminated. By the way, in the program settings you can select “Single click opens resource” for those who are unaccustomed to double-clicking to open a “subfolder”.

Noticed advantages.
Quickly characterize dynamic drivers and passive components.

Flaws.
1. "Capriciousness" in setting. Or am I missing something? I have little experience: a dozen setup attempts, and I don’t like the results of nine of them. Although, it is quite possible to take the characteristics of the DG, since they are needed only for preliminary calculations.
2. Narrow measurement range. Resistors - from 1 to 50 Ohms, capacitors from 0.05 to 100 µF (however, once the “box” guessed the capacitance of the capacitor to be 220 µF, but did not try to measure 1000 µF), inductances from about 10 µH to 3...4 mH , but more often only up to 1.5 mH.
3. Uncertainty in measuring the characteristics of coils with cores and incorrect measurement of the active resistance of the coils.
4. “I blinded you from what was”: actually AC
Don’t think that I spent more than six months only measuring the capacitance of capacitors. I also dug around with AS. Because of this, everything started... The task is to achieve maximum sound quality with the least amount of money spent.
4.1. Frame
My speakers had a difficult childhood. They were released around 1991 and served for about a year to provide sound in an open space on the market; sound cassettes were sold from the machine. So the hull got both sun and rain. Nothing was seriously warped, but the cracks became larger and the “self-adhesive” peeled off. The paint on the diffusers has faded and they have become dirty yellow. Then I'll paint it bright blue!
The body is made of 12 mm plywood, the front wall is 16 mm. Apparently, you can get a case made of chipboard.

But everything is fine inside. Maybe they really assembled the housing's seal, but there are no cracks inside. True, the “face” of the box is glued with cracks on the outside.

And here are the filters. In one speaker everything is fine. Only... in the mid and low frequency sections the coils seem to be the same, only some kind of piece of hardware is inserted into the low frequencies. Ah, so this is the core!!! But the filter is assembled according to the diagram

(50AC-106), and the capacity of the “Zobel” is as much as 60 µF, in the diagram it is less. But in the second speaker the midrange head was soldered incorrectly, in front of the quenching resistor. That's why the middle stuck out so much.
4.2. LF section
Let's look at the filter again. The coil has a core. We measure the inductance: 1.14 mH - clearly not enough. We take out the core and measure this L1 and L3 without the core (according to my diagram), select from them the one with the greater inductance, insert the core into it. It turned out to be almost 1.5 mH. Already better. We note that the inductance depends on the position of the core - it is smaller in size than the hole and was secured with a piece of foam rubber. What happens if the core is replaced? There was a steel wire lying on the balcony. I cut it into pieces of suitable length and filled the hole in the core with them. The result is already 1.8 mH. Not enough either. We remove the frequency response of the filter.

This is already better, only the filter setting frequency is not 500, but about 600 Hz. Apparently, it can be corrected only by winding up a little wire, which I did, again, I had a piece at hand, it turned out to be almost two layers, the inductance is 2.4 mH. Just what you need.

Well, the heads themselves. I was unlucky - the dust caps are made of plastic, so shiny. I'll change it to paper. In the meantime, I sanded the diffuser on one DG with sandpaper - it was wounded, but not completely. I washed off some black stuff on the outside of the rubber hangers with a solvent, then the glue and the same stuff on the back of the hangers. The result is equal suspension stiffness around the entire circumference. It is noticeable that the suspension of one DG is even softer to the touch than that of the second. And the DC resistance is very different: 7.9 ohms for one head and 6.8 for the other. And resonant frequencies are 35 and 43 Hz. Nightmare. The spread of other parameters will be very noticeable, which is not good at all (for example, Vas differs by almost half). After listening, a lot of overtones were identified in the suspension. After sizing once with "varnish" from the old "Moment" diluted in gasoline, it became better, after twice - almost good, you can live.

Since the heads are not new at all, “warming up” with a signal with a frequency of 50 Hz for half a day did not give any change in the parameters.

Instead of the original one, I tried to glue a dustproof cap made of thick polyethylene from a bottle of shampoo; it fits in size and convexity. The resonant frequency and output are slightly reduced and the frequency response at frequencies above 3 kHz is improved. But I peeled it off - it’s ugly, it already looks so-so... But this experiment gave rise to thoughts about the possibility of expanding the range of reproduced frequencies upward.

That's all for now. The plans are to replace the dustproof cap with a paper one, smaller in diameter and rigid enough to level the frequency response to 5 kHz or higher, this is important for a two-strip in order to abandon the use of 20GDS-2.
Everything was done with the body as usual: the joints were coated with window putty and covered with PVA with a layer of material from an overcoat and a layer of fairly dense padding polyester; a construction stapler is very convenient for attaching the “pie”. The free space is filled with cotton wool and batting. I note that there was no fabulous improvement in the reproduction of low sound frequencies, it just became a little better.

4.3. Midrange section
I washed them (they weren’t very dirty anyway). I made a PAS from padding polyester. One layer is not enough, two are needed, despite the fact that the padding polyester is quite dense. I trimmed the edges of the plastic glass so that the DG would fit tightly, cut out a gasket from soft linoleum, and filled the glass with cotton wool. It seemed to feel better, the singers stopped constantly screaming.

Frequency response of the filter

"20GDS, Vega filter" - "native". It is not clear how all this was supposed to be coordinated with the LF and HF sections. Bandwidth (level –3 dB) 1200…3500 Hz. That is, in my opinion, there is a guarantee of dips for a whole octave from 600 Hz, and from 3500 to 5500 Hz. And the “top” is almost not cut. The inductance of the coil in the filter is clearly high, the capacitance of the capacitor is insufficient, and it is necessary to supplement the filter with a Zobel circuit in order for it to work effectively.

The capacitance in the end turned out to be almost 25 µF, the coil is from the low-frequency section, the Zobel circuit is made of a resistor of about 10 Ohms and a capacitor of 10 µF (if it is possible, as I have in one speaker, to “pinch off” from a similar circuit of the low-frequency section, if not - take the one that is in series with the coil, you still need 25 uF there). Perhaps someone will like the sound of speakers with C2 at 15 uF, as V. Shorov advises - try it...

If you still want to replace 20GDS-4 with 6GDSH-5, use the experience of Dev-a (http://dev.azz.ru). It's a pity that the sound of the 6GDSH-5 does not match its younger brother 1GD-8, which has the same standard size. Even in a slightly “wounded” and healed state, this baby sounds lighter, more detailed and generally more pleasant, especially at low volumes.

In fact, an inductor and a Zobel circuit with 20GDS-4 are not needed, because this DG has a natural output rolloff above 4.5 kHz. V. Shorov wrote about this; in its predecessors - 25AS-2, 25AS-109 there was no coil. So we throw away the coil.

4.4. HF section
10GDV-2 (10GD-35 (B) in the old way) has not been praised by anyone for 25 years, but is still in use.
I didn’t change the fleece under the diffuser, but threw it out and stuck on a piece of Guerlain. You don't know what Guerlain is? I found a new link on the forum: http://www.sovmat.ru/material/m175.htm. If you believe Aldoshina’s publications, it should have good sound and vibration absorption at frequencies from 1...2 kHz, so, in my opinion, it is best suited for preventing overtones that arise due to reflection from the core of the magnetic system. In addition, there is positive experience in using Guerlain for gluing the internal surfaces of 2GD-36 and 3GD-31 head baskets. In my opinion, 10GDV-2 is the place for Guerlain. What happened?

The resonant frequency has decreased, which is always good. Many small peaks on the graph above 4 kHz have been replaced by one at 5 kHz, but larger ones may have to be dealt with. What did this do to your ears? It seems that the sound has become softer and even with a first order filter the well-known unpleasant “hiss” is not noticeable.

Now we go through the filter, making order II from III. What does this give? The small resistance of the inductor below the filter tuning frequency slightly “strangles” the DG output at resonance frequencies; in addition, by adding another capacitor to the filter, you can get a resonant circuit, tune it to the main resonance frequency (2 kHz), thereby further weakening the unpleasant overtones at the resonant frequency of the DG, although, it seems to me, this is already unnecessary, 10GDV-2 coated with Guerlain works like that.