What amp modules for compression drivers (hifi use)

[fb]

My speakers are AE TD15M-A, with BMS4550 compression drivers. Currently actively crossed over (via PC) and multi-amped using a home theater receiver.

I'm looking to move to a hardware DSP solution, and discrete amplifiers for sub/mid/high.

I need to find an amp module for the compression drivers. My criteria:

-Low noise/hiss - I often sit at computer within 2m of the speakers.
-Low idle power


I'm currently thinking of the miniAMP from minidsp.com, particularly as I can use their DSP as well.
It's TAS5704 based, 2x20W 8ohm BTL mode, 4x10W 4ohm SE mode.
TAS5704 datasheet says output noise is 89uV. Is this much?

Would it be wise to combine the miniAMP with a passive solution, ie. l-pad? I don't want to kill too much of the CD's sensitivity, particularly as I am EQing up the top end.

Any suggestions? I'm open to non-classD modules as well.


Thanks

[kstrain]

It is possible to make some low gain low noise amps with the LM3886. If you know a bit of electronics you can make a circuit with your chosen low gain (e.g. a few) up to 20kHz, but the necessary gain >10 for stability at HF. All it takes is to exchange the gain setting resistor on the "denominator" (i.e R1 in G=-Rf/R1 for inverting or G= 1+ Rf/R1 for non-inverting) with a larger resistor R1 in parallel with a series R C. R1 is chosen to set the LF gain, R to set the HF gain (must be less than 1/10th of Rf) and C chosen to give the appropriate corner frequencies (you want the slope to be between a few 10s and a few hundreds of kHz). Obviously sensible values of Rf etc must be chosen.

This trick ensures stability as only near unity gain in the feedback loop does the actual gain and phase matter.

If you use a nice hum-free layout you can get very high SNR. Depending on the horn or waveguide, it is still worth adding some passive EQ (e.g. a single capacitor high pass for a near-CD horn, and/or RLC notches for the peaks).

As you probably only need a few W (to deafen yourself at 2m) the power supply and heatsink can be small: choose a supply voltage a few V above the peak output you want (I think I used +/-20V)

I used neat little PCBs bought from "Mad about sound", some very slight bodging allows the required layout to be achieved reasonably neatly.

When I was driving a BMS driver on a 18sound horn from a DCX2496, I used a gain of just 2, which was plenty.

The same arrangement will work with other (non class D) chips, with small variations in the detail.

Ken

ps. I deliberately did not give the values of Rs and Cs, as if you can't calculate them I suggest not trying this trick.

[jerome69]

See also TDA 7295

[fb]

I looked at the datasheets for both LM3886 and TDA7295 - is the "Input Noise" measurement in these datasheets equivalent to the "Output Noise" measurement I see on the datasheets for other chips? It's the only measurement I see in uV but it's MUCH lower than other chips.


Would it be possible to use a LM3886 standard pre-built module, or kit, instead of building a custom one? I assume the output noise is independent of available output power - it doesn't bother me if I have more power available than I need.




Does anyone know how to calculate an LPad value based on an output noise measurement from a datasheet?


Thanks

[doug20]

fb, a couple of members on AVS have been explaining that we need to have amps with very good gain structures and we should be turning the gain structure on that amp down to maybe 1/4" that reduces the noise levels involved. Pre/pros can increase the voltage needed.

FWIW, The best amps S/N ratios I have found where in the Parasound amps. (No VCs on them though).

[fb]

Hey doug Thanks for that. Can you link me to a post/thread? I'm not sure that the gain of the amp has anything to do with the amp's inherent noise. I've tested both my XR50 receiver and a TAS5630 module with nothing connected on the inputs and the noise is the same.

[doug20]

There has been difference discussion, centered around using the DCX. One post from Steve71 is here..
The semi clone of JBL 3731's build thread - Page 13 - AVS Forum

You might want to PM him to get a better explanation.

[kstrain]

Originally Posted by fb

I looked at the datasheets for both LM3886 and TDA7295 - is the "Input Noise" measurement in these datasheets equivalent to the "Output Noise" measurement I see on the datasheets for other chips? It's the only measurement I see in uV but it's MUCH lower than other chips.

Generally in these types of applications,

output noise = input noise x gain.

That is why I suggested a low gain (just what you need) amp. You can of course have lots of gain then pad down the output, but that means a finite resistive source driving the compression driver, and that in turn will affect any filtering you try to do after the amp. Also if you amplify up then pad down you are paying for a larger supply and (at least with class AB) a larger heatsink than you would otherwise need.

Most prefab modules have quite high gain (x20 or so).

Ken

[Brett]

Originally Posted by kstrain

Generally in these types of applications,

output noise = input noise x gain.

That is why I suggested a low gain (just what you need) amp. You can of course have lots of gain then pad down the output, but that means a finite resistive source driving the compression driver, and that in turn will affect any filtering you try to do after the amp. Also if you amplify up then pad down you are paying for a larger supply and (at least with class AB) a larger heatsink than you would otherwise need.

Most prefab modules have quite high gain (x20 or so).

Ken

So get the cheap 2ch kits off ebay and change a resistor or two/channel and set the gain where you need it.

[fb]

I read that LM3886 requires a gain of between 20 and 30, so perhaps I would need to use the trick kstrain described above.

Trouble is, I don't want to limit myself to 2 or 3 watts - these do get taken outside sometimes, and I'm using high freq EQ.



I'm trying to work out how low I actually need the output noise to be. Using the measurement from the TAS5630 datasheet of 270uV, and an 8ohm, 113db/W/m driver, I get:

P=(270*10^-6)^2/8
SPL=113 + 3 * Log2(P)
SPL=32.87dB at 1 metre

Is my maths correct?

This was quite loud, at least at night when I tried it.