Advice for DIY fixed ratio line level attenuator (voltage divider) for active DIY speaker

[Shefffield]

Hello forum,

Not sure if I am in the right section or should have posted in DIY...

I need some help in designing and building some fixed ratio attenuators for my speaker setup. Digitally controlled active speaker, each driver on its dedicated DAC and amp channel.

Recently I damaged a tweeter by accidentally putting out full power after a mis-click in Totalmix FX. It is possible to deactivate the functionality "double click toggles -infinity dB/zero dB", but after a Windows crash this somehow got deactivated. Scratch one ribbon tweeter, thanks Microsoft!

So my challenge is to solder some attenuators to limit the maximum level I can torture my drivers with. It also won't hurt to not use too much digital headroom for the volume control and to be able to keep the virtual fader around -10..-30 dB instead of -50..-70, to pick just any numbers.

I am still learning electronics and would feel much more confident in actually building and connecting anything after some checking by you.

Signal source:

RME Fireface UFX II

Fireface UFX II - RME Audio Interfaces | Format Converters | Preamps | Network Audio & MADI Solutions

RME Audio Professional Live, Studio, Recording and Broadcast Solutions. Unrivalled Quality, Performance & Stability MADI Interfaces, Converters & Preamps.

rme-audio.com

DA Line Out 3-8 switchable:
+4 dBu: 1,227 V RMS / 3,472 V peak to peak
-10 dBV: 0,316 V RMS / 0,894 V peak to peak
Maximum output level: +19 dBu (6,903 V RMS / 19,526 V p-p)
Output impedance: 75 Ohm
Connection: 6,3 mm TRS jack, servo-balanced

(Calculator used for the conversion to volts: http://www.sengpielaudio.com/calculator-db-volt.htm)

Signal sink:

Powersoft D-Cell 504

https://www.powersoft.com/en/products/amp-modules/d-cell-504

Amplifier gain: 32 dB (x40 voltage gain)
Input sensitivity @ 8 Ohm: 0,94 V / 1,7 dBu
Maximum input level: 4,36 V / 15 dBu
Input impedance: 10 kiloOhm balanced

First question: T-pad or H-pad?

I found these schematics and calculators:

T-Pad / H-Pad Calculator - Resistor Values for Attenuation

Voltage Divider Calculator

mh-audio.nl

I'm going to use balanced connections, so I'll have to make identical attenuator circuits for 2 and 3, right?

I think the most elegant solution is to build a few attenuators in plugs, with different fixed levels, so I can run my tests and measurements and chose the matching attenuation by simply swapping plugs. However, TRS plugs needed for the UFX II don't have any spare room, and the D-Call modules have shrouded header connectors. Any recommendations for inexpensive, but reliable connections I could use here?

Thank you in advance,
Axel

 

[AnalogSteph]

First question: T-pad or H-pad?

Neither - double L. (Audio circuits are not impedance-matched, and you don't need the attenuator to work in both directions.)

R1 = R3, R2 = R4, match resistors by hand (plain Yageo 1/4 W 1% metal film should do fine). Use heatshrink so you can fold the whole contraption like a mirrored Σ that got squashed flat.

You want +19 dBu --> +2 dBu, so ~17 dB of attenuation. Shooting for a similar input impedance again, you could try:
R1 = 4.7k, R2 = 910R
R1 = 5.1k, R2 = 1k
R1 = 4.3k, R1 = 820R
(all ca. -17.15 dB, input impedance between 5.0 and 5.9 kOhms x2 = 10.0 and 11.8 kOhms)
or if you insist on using E6 values that you may already have on hand,
R1 = 4.7k, R2 = 1k
should do as well (-16.56 dB, 11 kOhms).

In addition, I would research the feasibility of further passive attenuation for the tweeters, increasing output gain in the crossover to compensate (mind your headroom, of course). Since we are talking Class D amplification, you'd be using an L-pad using appropriately rated resistors to keep input impedance in the correct range.
If e.g. you wanted to restrict max tweeter power input to <10 W on an 8 ohm tweeter, you could try R1 = 5.6R (100 W, maybe 2x2 of 25 W each in series+parallel), R2 = 2.2R (25 W) for a total attenuation of -12.7 dB and input impedance of 7.3 ohms. Since the module is still supposed to work fine at 16 ohms (minus some FR deviation at the top end I assume), 12 ohms (ca. 60 W) + 6.8 ohms (ca. 20 W) may be a more attractive proposition in terms of power dissipation and parts cost.
Or if up to 20 W would still be OK, 3x 3.3R (20-25 W) in series + 10R (20-25 W) for ~10.2 dB at ~14 ohms in would do the trick as well.

 

[Lambda]

i would attenuate without ground connection

 

[AnalogSteph]

i would attenuate without ground connection

Note that the floating U-pad provides no attenuation of common mode to speak of. It has fewer resistor matching issues than the double-L, but in this day and age and in a DIY project that's not really an argument.

In something like this it's probably a wash, but if you really need a lot of attenuation (e.g. line to mic level) and have a lot of CM to deal with (e.g. an impedance-balanced output), the double-L seems like the safer bet.

 

[Shefffield]

Thank you all, guys! Very helpful!

I'm currently traveling and will return only next year. Then it's time to heat up the soldering iron.

I think it would be a good idea to place the L-pad at the end of the cable running from the DAC to the amp (in the plug connecting the amp module then), so the voltage is still high over the long distance. Or does this not matter?

The idea of reducing the max power behind the amp is interesting. I could only get 4 ohm tweeters though. Aiming for very small physical size, so the choice is very limited. The good thing is that the tweeters I currently experiment with are not so expensive to replace in the worst case. Yet.

Limiting to 10 W seems a bit extreme though, what about transients? I've seen reviews of common dome tweeters that survived bursts of 50-80 W without much trouble. I know, the 20 mm small VCs would overheat rapidly if that was sustained, but for jut a few fractions of a second?

I don't expect resistor matching to be an issue. Low tolerance resistors are not expensive, even 0,01 % should not break the bank, if I'm not mistaken. Any small deviations will be addressed by the final digital correction anyway, so it's not really critical, as long as it doesn't change over time.