cascade filters What resistance should I change to reduce the gain?


2016-08-23 12:58 am
I think you're looking at the single 5532 near the input there, which seems to be an input amplifier (surprise!).

If I have deciphered the resistor color codes correctly (stupid things are always a bugger to read), the input circuitry seems to be:
100R in series, 56k to ground, some capacitor to ground.
First 5532 seems to be a noninverting amplifier, 390R / 100R. They may have expected rather low input levels?

Please verify. It's quite hard to distinguish a 100R 1% (brown - black - black - black - brown) from a 1k 1% (brown - black - black - brown - brown).

What sort of levels are you looking to run into the thing, and what do you need output wise?

If you are lucky, you might get away with removing the 100R in the input amp, making it a unity gain stage - that's about 14 dB less gain.

Please do verify that output noise is low enough once you get that far (input shorted). Given the circuit designer's apparent penchant for low feedback network resistor values, I almost assume that this is going to be the case, but if you have a low-noise power amp and sensitive drivers you might still need an extra output attenuator, maybe on the power amp side. You need to be matching driver levels somehow anyway.

If you get too much noise in the treble section from just the power amp alone, consider a passive L-pad attenuator. There are too many active speakers with plainly audible hiss already.

I hope you have the kind of speaker measurement setup to make all of this work. The drivers pretty much have to be flat in frequency response and perfectly matched in amplitude and phase / delay around the crossover points for such a "generic" crossover to work well.


2016-08-23 12:58 am
thank you very much for such an extended response

I will eliminate that resistance

the drivers are beyma tpl150 and beyma 6mcf200 both with a sensitivity of 100 db I think I could use a unit gain.

if I have measurement system and equalize by means of apo.
the drivers are beyma tpl150 and beyma 6mcf200 both with a sensitivity of 100 db I think I could use a unit gain.

In fact, this may not even be sufficient.

At 100 dB/2.83V/m, for the output noise to equal 0 dB SPL it would have to be at 100 dB below 2.83 V, or 28.3 µV.

With a power amp gain of 26 dB, this requires a total input noise (power amp + preceding stages) of 1.41 µV.

Problem #1 - a typical chipamp like the LM3886 is at 2.0 µV (typical) all by itself, that would mean ~56 µV out. A good discrete power amp might manage 30 µV out. The circuits usually have to be run at a minimum gain of 20 dB, and may be a bit tricky to stabilize even at this point.
EEVblog #1128 - Studio Monitor Speaker Noise - Part 2

Note: Noise at power amp input level will be reproduced by the connected driver across its entire native bandwidth. For the 6MCF200Nd, that's about 400-500 Hz to 10+ kHz. This will add to tweeter noise, especially on axis (another ~3 dB at ~1.3 kHz up).

(BTW, the 6MCF200Nd rolls off below ~500 Hz on the infinite baffle, how do you expect to be crossing it over at 310 Hz? This will require some crossover mods.)

Problem #2 - the crossover is not completely noiseless by itself for obvious reasons. Even completely disregarding any resistor noise, a 5532 at unity gain will be contributing at least 0.7 µV across the audio bandwidth, and here you need two in series, so it'll be 1.0 µV, or +26 dB = 20 µV. Plus some overlap between frequency ranges.

In sum, you would be hard-pressed to get better than anywhere between 36 µV and 59 µV, or 2-6 dB SPL @ 1 m. Probably worse due to overlap. Now I found noise at about 7 dB SPL to be not overly disturbing but definitely audible, so 6 dB would probably be OK and 2 dB bordering on vanishing.

Problem #3 - DAC outputs have some noise, too. A more than decent 111 dB of dynamic range at 2 Vrms means a whopping 5.6 µV - that would be dominating everything else here! (12 dB SPL @ 1 m all by itself.) You would have to get to the 126 dB (unweighted) mark to get down to 1 µV. You can definitely buy DACs that meet this spec but they would still count as high performance.

So yeah, this is getting silly. In this scenario, your crossover signal level at 100 dB SPL would be all but 141 mVrms.

Now what sort of application do you have in mind? In home hi-fi, 110 dB / 1 m is probably the maximum you are likely to go for. That would be a whopping 447 mVrms in your crossover (and at the DAC output). Not sure what a standard crossover level would be, but I think 1-2 Vrms max should not be too much to ask for. So chances are you would need a total power amplifier gain 7 to 13 dB less than 26 dB, i.e. 13-19 dB.

This may require special compensation tricks such as draining away open-loop gain to keep a circuit stable below specified minimum stable closed-loop gain, or a different kind of power amplifier altogether that has more in common with a headphone amp.

At the very least, a passive attenuator between the DAC output and crossover input should at least enable you to use a "normal quality" DAC. 6.8k in series and 2.2k in parallel or so. You could integrate this onto the crossover board, otherwise you might as well just use a 10k volume pot.

Interesting, all of this. I was basically aware of these issues but had never put all the calculations together so far. Designing active systems with high-sensitivity drivers definitely has a few pitfalls to offer!

it seems to be 1k
Which one, the resistor going to ground next to the input opamp? Sure looked like 100R in the photo. What's the one next to it then - 3.9k? (You can remove the opamp temporarily.)


2016-08-23 12:58 am
Right now I have the beyma connected to this headphone amplifier

GZLOZONE terminado LJM HA PRO2 Nivel de monitor de amplificador de auriculares L11 32-in Amplificador de auriculares from Productos electronicos on | Alibaba Group

the sound is much better than with any other amplifier

I have ordered a couple of these for the treble and medium(it would be nice to be able to put them in parallel):

AIYIMA TPA6120A2 amplificador de auriculares de alta fidelidad Tarjeta de amplificador de Audio de fiebre afilado Imperial de Athens amplificador de auriculares-in Amplificador de auriculares from Productos electronicos on | Alibaba Group
I listen to 1 m from the speakers so I think that even without putting them in parallel they would be plenty at that distance.

The crossover is customized

It is 4 ways with cuts in 140 hz 700 hz and 2200 hz

I have some lm4562 ... when placing them it sounds much worse, surely I have to modify the whole circuit so that they sound like they really should sound

Last edited:
Hmm. According to here that's an NJM4580 inverting driving three 4580s as buffers combined with 10 ohms each. That'll get you a nominal damping factor of 2.4 re: 8 ohms - not great and potentially quite "tubey". And while it should have a decent amount of oomph for 32 ohm headphones, I still can't imagine it would be too happy with an 8 ohm speaker load... you might get 1.7 Vrms out with a bit of a following wind (or ~350 mW), and we better not talk about distortion at this point (or opamp case temperature). Mind you, with these drivers that's still about 95 dB SPL!

Noise should not be an issue with this one though.

Still, this is rather bordering on abuse.
Note: TPA6120A2 based amps often have 10 ohm output series resistors for stability reasons, and this one I think is no exception. Not ideal in your case. Damping factor 0.8, 1/4 W resistors no less. (They do not tend to be excessively happy driving <60 ohms either, though the presence of a decent heatsink gives some hope.)

I mean, I'm all for current driving but only if you can keep the frequency response in check somehow.
The crossover is customized

It is 4 ways with cuts in 140 hz 700 hz and 2200 hz
So, what do you need this 3-way crossover board for then, the one that prompted this thread? :confused:
Last edited:


2016-08-23 12:58 am
It is strange but the small amplifier does not heat anything ... and the distortion to the ear is zero and much lower than with any other amplifier due to its low gain.
I have two cascade filters, the two are the same, only one is 2 and the other 3-way, the seller gives you the option to configure the custom cut hehe