Please help with sound quality- op-Amps

[Nora]

I need some advice and help with sound quality. I'm totally new to this forum and did do some searching before posting but not totally sure what I'd be searching for....
The idea is that a recorded at 5kHz, 8 bit, mono wav file is stored in an eeprom as a hex file, then goes thru a DAC (signal comes out as current for this particular chip) to a current2voltage converter (LM358 op-Amp) with a high pass filter that passes everything above 20Hz, then another gain stage (LM358), then an Audio op-Ap (LM356). The software is programmed into a PIC chip.
I attached the schematic.
The trouble is that the circuit/software works, but sound quality is very poor. You can hear the recorded file, and while you can understand what it says, and it does not seem to be terribly distorted, it is corrupted with lots of noise.
This is for a school project and I have no choice about the memory chip and process (altho I am open to suggestions). What I really need help with is filtering and noise control. I don't have to use those op-Amps.
Another question: is it better to use BJTs than op-Amps for sound?
Thank you!
Sophi

[lineup]

LM358 can be exchanged for other OP-Amp.
For this project maybe a TL072 would be a good choice.

But, I am not convinced OP-Amp LM358 is responsable
for the poor sound quality in this circuit.

Only 8 bits at 5 kHz is not top of digital storage of sound.
Maybe you can make a better filter in first OP-Amp.

There are members here that have threads about such Digital to Analog Conversion circuits (DAC)
and I/V-filters they have built and developed.
Maybe you should have a look and ask in Digital forum

[poobah]

Nora,

If you sampling rate is 5 kHz... then your low pass filter after the DAC should be set at a maximum frequency of 2.5 kHz. This is called the Nyquist therem or something like that. Your filter should also be "multiple order", at least 2nd order (8th order would better), to really kill the hash in the higher frequencies...

I see you have 2 op amps there... your filters could be incorporated "around" thos amps very easily.

Google on"2nd order lowpass op-amp" and you find lots of circuits... you want a "Butterworth" for simplicity You can turn each amp into a 2nd order to achieve a 4th order. This will give you a vast improvement.

Set your "cutoff" frequency between 2 and 2.5 kHz and punch through the equations.

If you can't find a circuit, lemme know...

[richie00boy]

Yes, and seeing as voice is typically 300Hz to 3.3kHz as defined by telephone bandwidth, you see that a 2.5kHz limit is going to give you problems.

[poobah]

Yeah... but it'll work for a grade...

[AndrewT]

Hi,
the virtual earth (inverting) opamp has no gain setting resistor in the input line. Is this normal for a I2V convertor?
Where is the 20Hz high pass filter?

Are you the pupil or the teacher?

Full marks for resourcefullness but zero marks for independant research and/or problem solving.

[Nora]

Hi-
Thanks for all the replies.
Um..what I already have will work for a grade....I want to make it better!
Would I want a low pass or a band pass filter (I am going to look at Butterworth).
Is this right- sampling frequency (mine- 5kHz) is 2x bandwidth and bandwidth is the frequency at which I can get sound (needs to be minimum of 3kHz for clarity).
So I should design my filter around 3kHz and record at sampling rate of 6kHz? BTW 6kHz is the fastest I can have with my Micro Processor speed.
Sophi

[Nora]

Andrew-
I am the pupil.
The op-Amp directly after the DAC is set up as a current2voltage converter....it takes a negative (backwards) current at pin 2 input and converts it to a positive voltage pin 1 output. So no need for a resistor.
The other part of the op-Amp is a non-inverting op-Amp.
Hope that clears things up for you-
Sophi

[Nora]

Oh, the 1000uF right before the speaker is the 20Hz filter. I calculated wrong?
Sophi

[Eva]

The filter attached to the first op-amp (4.7 k || 470pF) is actually a low-pass filter, not an high-pass one as you have mentioned. The basic math (f=1/(2*pi*r*c)) shows that the cutoff frequency of that filter is set whay too high (almost 70Khz), and this is your main problem.

I would use a sharp filter of at least 4th order, manually tuned to provide at the same time minimum attenuation at 2.5Khz and maximum attenuation at 5Khz and above. Note that this noise that you hear through the loudspeaker is nothing but aliasing components from 5Khz upwards, in fact, everything coming out from the DAC at 5Khz and up is just trash and should be discarded.

If you have some unused processing power in your PIC, you may consider operating the DAC at 20Khz and fillng the three new samples created between each 5Khz sample with linearly interpolated values. You may consider second order interpolation, altough I think that it would be prohibitive in processing power terms. But, why interpolating at 20Khz? Because it will reduce dramatically the amount of 'trash' coming out of the DAC between 5Khz and 20Khz (it will be actually placed at 20Khz and above), making filtering much easier. (By the way, this technique is called 'oversampling', by a 4x factor).