Good chip amp for driver measurements.

[Aucosticraft]

I need a chip amp, for general purpose, that would be used while testing new speakers or measuring frequency response of new drivers or for the breakin of drivers.

So I am expecting power rating of 10W @8E or more.

Good THD rating, flat response between 10Hz to 30KHz.

At first thought I had LM3886 in my mind but not sure if its worth. Also suggest any other that can be used.


EDIT: I just saw LM1875, Is it good enough? would you suggest it ?

 

[JMFahey]

In a rising power scale: TDA2030 LM1875/TDA2050 LM3886 TDA7294

They are all very good and meet or exceed your specification, I'd use LM3886

Chip cost is nil, compared to PCB/PSU/transformer/cabinet/hardware .

 

[PetruV]

tda2003 if you use single supply

 

[johnr66]

tda2003 if you use single supply

Poor advice. You would have to use two TDA2003 ICs in a bridge design and operate near the max Vs. to get the 10 clean watts of output into an 8 ohm load.

TDA2050/LM1875 is the better suggestion. You may need to adjust some values of components to be sure you are getting the full response range you need.

 

[ChristianThomas]

Personally I would use the LM3875 over the 3886 unless you want high current delivery. It doesn't have a ground on the chip and that makes for fewer mistakes and an easier layout.

 

[Aucosticraft]

Thanks friends,

After reading you all I think I should go for LM3875. as this amp should serve my purpose.

Higher rating amp would add cost of powersupply too. I am still not sure if I can get flat response from this chip in frequencyband of 10Hz to 30Khz. ( not about the input filters)

If any body has a idea, please share.

 

[lanchile]

Thanks friends,

After reading you all I think I should go for LM3875. as this amp should serve my purpose.

Higher rating amp would add cost of powersupply too. I am still not sure if I can get flat response from this chip in frequencyband of 10Hz to 30Khz. ( not about the input filters)

If any body has a idea, please share.

"if I had to build a chip amp" I would build an LM3875. but, Why don't you try a discrete amp class A or class A/B!

 

[ChristianThomas]

Thanks friends,

After reading you all I think I should go for LM3875. as this amp should serve my purpose.

I am still not sure if I can get flat response from this chip in frequencyband of 10Hz to 30Khz. ( not about the input filters)

If any body has a idea, please share.

30kHz - in its stride! In your application I would probably put a second order low pass Butterworth at a frequency above this so you know it's flat in your band of interest, rather than have it degrade by itself. Say at 80k. That is, if -0.2dB at 20kHz is too much off flat for your purposes - which is probably what you would get with a single pole. It would be a good microphone though, to make this important.

 

[Aucosticraft]

"if I had to build a chip amp" I would build an LM3875. but, Why don't you try a discrete amp class A or class A/B!

discrete amp could have been my choice but, I wish to keep it simple and easy to repair ,if in case.

30kHz - in its stride! In your application I would probably put a second order low pass Butterworth at a frequency above this so you know it's flat in your band of interest, rather than have it degrade by itself. Say at 80k. That is, if -0.2dB at 20kHz is too much off flat for your purposes - which is probably what you would get with a single pole. It would be a good microphone though, to make this important.

Well, I believe LM3875 don't need compensation if used as shown in reference application. I am not sure about the phase errors. But I hope it would perform flat.
Adding extra RC filter will disturb the phase depending on frequency. And this is not good for measurements.

Correct? So, I thought I would avoid any input filter. And hope it would be stable, and flat in the given frequency response also, would have negligible phase error.

 

[ChristianThomas]

Well, I believe LM3875 don't need compensation if used as shown in reference application. I am not sure about the phase errors. But I hope it would perform flat.
Adding extra RC filter will disturb the phase depending on frequency. And this is not good for measurements.

Correct? So, I thought I would avoid any input filter. And hope it would be stable, and flat in the given frequency response also, would have negligible phase error.

No, it doesn't need compensation, though an L//R and a Boucherot cell are a good idea, and can be lightly done.

I did mean having an input filter, which is something I would do largely to define the shape. (It would be a bit risky to trying and do it on the amp itself, I suspect.) But I don't think you see the effects of the amp's own parasitics until beyond 200k, so yes it would be absolutely flat. Certainly flatter than many audio power amps. (And that means phase, too.)

I can't see any reason why you would need (or want) a discrete design. You mostly get distortion of around 0.02% or below and you also have much more open loop gain to play with than you could probably find easily by yourself.

 

[Arty]

if you choose LM3875, then go for the TF variant. (isolated from heatsink).

allso, Your problems are not going to be the chipamp and its phase shape.
But, getting a proper microphone, and a proper microphone pre-amp, and even something harder,
and enviroment suitable for mesurements are the real problem.
That is, if You are after freqvency response curves.
Actualy room modes, rezonating objects, room gain, reflections and many many other things will have larger impact.
and here is a hint, there are amny software that can actualy use the soundcard intput and output jacks to calibrate it self.
The idea is that is will generate a sine sweep , that goes into Your amp. The amp output is connected to the input of the soundacrd. The difference between the recorded and playd sounds are doe by the amp. So from then correction can be done.
Naturaly before this the input and output are connected with no amp in between to calibrate the stuff.

allso, if, that IF you know what You are doing, and do not have just a verry tiny DC offset You can omitt input caps on an amplifier. So it should go down near DC.
Upwards.. well most chip amps can actually go verry high. Even if they could not, software can correct the falling response curves to some degree.
I'm pretty sure LM3875 TF will be more than enough.
Usually a simple TDA 2050 is more than sufficient for the job.

 

[ChristianThomas]

allso, if, that IF you know what You are doing, and do not have just a verry tiny DC offset You can omitt input caps on an amplifier. So it should go down near DC.

Is it worth it, for the risk of blowing up something (next gen Revelator perhaps?). 10uF and 10k gives you less than 2Hz, which should be low enough.

 

[AndrewT]

10uF & 10k give 100ms RC time constant.
For reproducing music and LF sound effects to go along with some music then 100ms is sufficient for all very wideband speakers.

For measurement purposes I would guess one needs to increase that LF bandwidth by a decade to remove the filter effect from the measurements.
DC ability is probably not required but maybe an RC ~ 1s would give a good measurement system.

 

[Arty]

if you know what you are doing then it is worth it.
if not, then don't even thinker with it.

 

[Aucosticraft]

I agree microphone, and preamp contributes a lot to the improper measurements. So to minimize this, I have got the Preamp calibrated. For mic I have calibration file, which software uses for adjusting the errors. Despite of all the care taken, there is going to be some error creeping in. But I just tried to keep errors the lowest possible.

Regarding using LPF, I don't wish to use one but, would use it for safety against accidental offset DC ( wont know if all players connected while listening would have zero DC offset).

I feel 0.1Hz LPF should solve my purpose of near flat response and zero phase shift in the required frequency range .I have attached the screenshot pls view, and comment.

 

[ChristianThomas]

I'm not sure how much protection 0.1Hz is going to afford you. What happens when you connect DC to the input? The capacitor charges for a few seconds.

I really don't think you need go down that far for anything to do with loudspeakers. At LF one generally infers the response from the parameters, which you can measure. And the speakers themselves are producing nothing to speak of at 5 or 10Hz - perhaps 30 or 40dB down.

You may want to measure around the LF resonance to see if it is doing what you expect and nothing in the accuracy of this will be affected by a pole at 1.6Hz. A different temperature in the room or of the voice coil will have a far bigger effect and, swamping them all, the amplitude at which you take the measurement will have the biggest effect of the lot, moving the resonance a semitone or two.

 

[JMFahey]

2Hz HPF is more than good enough for all the loudspeaker measurements you may ever make.

Any lower, only for earthquakes, Nuclear test blasts and whale communications.

 

[Aucosticraft]

OK, so will be using 2Hz. HPF & will make a provision to bypass the filter. and compare the difference that can be seen in real life measurement.

 

[Arty]

that seems like the ideal way, yes.