DIY Audio Projects Forum

Hey,

i just wanted to put my cheese on it...

This is an excellent work. Absolutely great! No more words are neeed here.

Just a little tip -
when you go to put it on the board, try to keep the length of the tracks from the input resistors to the IC input as short as possible.
Also the length of the feedback resistors path at the IC pins should be short as possible. Long tracks or leads / wires after the input damper resistors
(the resistors connected from the input jacks to the IC inputs) often tend to pick up noise and High Frequency (Radio) influences.
Keeping them short also prevents possible oscillations. OK - the OPAs are very stable, but - the better done is the it better works.
You can run miles of wires (ok - not really miles) before the resistors, but after them it's going to be serious...
Studio equipment often uses HF suppressors and gate resistors at the signal inputs (input signal conditioning) or HR / RF dampeners.
The 10K input resistors already are "gate resistors" and should keep the inputs clean - when the path is not too long....

Have a nice Weekend .

_________________
Cheers,
Tom.

Some of my projects: TDA2050 Chip Amp, the LM3886 Gainclone Thread and the Szekeres Headamp Thread.

For mounting, i was looking at having the resistors close as they can get. With some slight mods, an off the shelf board should work just fine.

This board is made for + and -, but with some trace cuts will give me a perfect VCC VGND and VEE(more cuts can give the audio its own ground to only join the PSU at one point as well). It also offers room to either use jumper wire OR input filtering caps before the summing resistors(I have a feeling I will be looking to add such caps for the sake of finding a sound card that DID push some dc). A set of sockets would allow easy swapping of caps to test.


Any recommendation on the dc blocking caps would be appreciated. I even have a sound card that seems to have higher value caps(470uf I think) and that seems strange.

Hey,

for the input decoupling caps i would go for MKP / MKS types. They are very common and widely available. Panasonic, Wima, Vishay - just to name some.
You have already have mounted them after the supply filter caps for example...
The higher the voltage the higher the pin grid. 63VDC are small to fit the board grid and should be enough. The right value is between 0.1uF and 0.47uF
which is 100nF or 470nF. Here comes a little problem to solve... The circuit doesn't include a leak resistor or signal conditioning / impedance matching.
There is no resistor that goes from the signal input to ground and therefore you have to regard the input impedance on the OPAs as real impedance.
Thanks god you go with the OPA2134 and they have a very high input impedance due to the FET characteristics. Win situation!
Our datasheet shows 10x13high II 5 per Ohm II pf which is nearly endless. Now the CR-Filter is formed by Cin and Rin(FET).
A 100nF Film capacitor in combination with the FET input (guesstimated average 100K minimum) provides a high-pass starting at 15.92Hz.
When this is not enough - 220nF or 470nF will be technical overkill... But - with every nF the time constant does rise. Will say that 0.1uF has a 0.01s delay
(taken from the sengpiegel audio online calculator). 0.47uF alreday delays 0.047s (3.39Hz). Thus 0.22uF should be the maximum for accurate results and lowest frequency cut-off.

A very good and always helpful calculator (for example - language button is on top of the page): http://www.sengpielaudio.com/Rechner-RCglied.htm

_________________
Cheers,
Tom.

Some of my projects: TDA2050 Chip Amp, the LM3886 Gainclone Thread and the Szekeres Headamp Thread.

If i see this right, with a variable resistor to ground and the right cap, I can make a filter to remove all that humming sound on some tv channels?

Is this also the basics for a equalizer.

On topic, it looks like a 1uf seems to provide fairly good results in that calc. Does a higher value effect the sound. I notice that many sound cards still use electrolytic caps(is this harming the sound, I see some people mod and bypassing/swapping them).

I just gave it a shot with the 0.1uf caps(before the summing resistors) I have for the power supply filtering(had extras)

Now I have a high pass filter(with some tweaks this may be useful for TV ).

Is it because I am in front of the 10k resistor so it acts like a high pass?, If that is the case, a 1uf gives me 16hz(Below my hearing I would guess) and a 2uf gives me about 8(Highly doubt my computer can go this low, but something in the future may).

I think i need a capacitor value pack or something to play with.

I am still using the op amp inverting. So i would have low impedance(so something like 10k resistor + opamp inverting input) right?

Thanks for that calc btw.

Not exactly. The frequency filter network is determined by the input capacitor and the real input impedance. With the simple input circuit - without resistors referenced to ground -
you have an input impedance of nearly infinite. This is very high and this high impedance does not provide a load for the input path. This is due to the FET input of the OPAs.
FETs in a single device amplifier (1 FET with some external components), FETs can have a lower or a high input impedance - depending on the amplification mode.
The OPA inputs are switched as Source-Followers (Common Drain mode) attached to current sources. Source followers usually have a higher input impedance,
which results in a low output impedance. (Did i mix it up in the earlier posts??) The FET inputs provide an uneven higher impedance compared to Transistors.
OK - i hope i don't mix it up now.. The scheme above should show the actual situation. You have a input capacitor, a 10K serial resistance -
(because it is in serial with the input) and the naked FET input. C1 and R2 are a serial network and the FET is the real impedance**. The FET impedance usually is 1M to 10M (100K worst situation guesstimated).
Now you can count the 0.1uF and the FET impedance (let's take worst situation values - 100K). This provides a 15.9Hz corner frequency - low enough to shake glass.
The serial resistance of 10K (R2) does not join the party - except that it attenuates the incoming signal in conjunction with the feedback resistor. Until here - the corner frequency is only determined by Cin (C1) and the FET -
Changing the corner frequency means changing the capacitance all the time.

To make the input corner frequency adjustable, you can go the simple way and add an optional linear (lin not log) potentiometer* (switched as variable resistor). This allows you to "bend" the frequency curve
like if you would use an equalizer for just the high-pass (low cut) band. This pot (variable resistor) is attached to ground and now works in parallel to the FET input.
The equation (simplified) now is fc = 1 // 2 * pi * Rpot II R_FET * Cin where R_FET is obsolete due to the infinite (best case) impedance.
Now the potentiometer becomes the main input impedance and when you change the resistor value by turning the potentiometer you get the adjustable fc (corner frequency). Puh.
It seems to be simple - but it isn't. As you can see, the potentiometer with the variable impedance is attached to ground. Every turn which cuts the low frequency does result in a lower input impedance and it does load the input.
More cut-off = more load == more load = more loss. It simply also attenuates the signal.
When we stick with the 0.1uF input cap and a 100k potentiometer we get the hum frequency cancellation at about 25k set at the potentiometer (63Hz).
25k is not even bad and the loss is not super high. 25k is still a good impedance value for usual input networks. Some go and mout 10k volume pots (which i never liked - personally one) as input volume regulators.
This can be seen as low. 100k might be OK so. As said - it is the simplest way and only a "Bass cut" (low cut) network.
Unless you don't need an active circuit (with loss compensation) it works. If you want to engage a "multi band" equalizer you don't come around to build an active EQ.
Every single band adjustement does result in heavy losses if you drive the bands without compensation. This is going to be complicated.

Some examples and schematics for further reading:
http://www.sentex.ca/~mec1995/tutorial/xtor/xtor5/xtor5.html Simple tone controls - scroll down to 3/4 of the page - good and useful infos
http://gilmore2.chem.northwestern.edu/projects/equal_prj.htm Baxandall controls - lossy unless you put another OPA inbetween - good and useful infos
http://www.free-circuit.com/3-band-graphic-equalizer-circuit-with-ic-tl072-and-ne5532/ let's get serious (OPA2134 also works)
http://www.free-circuit.com/10-band-graphic-stereo-equalizer-circuit-with-tl074/ if someone needs more channels... 20 for stereo

_________________
Cheers,
Tom.

Some of my projects: TDA2050 Chip Amp, the LM3886 Gainclone Thread and the Szekeres Headamp Thread.

I must have something wrong because when I used the same 0.1 uf capacitors used for HF filtering the power supply, I get a very harsh high pass filter.

If i had lets say a 100k resistor to ground after the capacitor(and before the 10k summing resistor), that would take the op amp out of the filtering equation allowing a cutoff at about 16hz?.

As it is now, i do not have access to a lot of capacitors and digikey orders all ship at the same price making it less then ideal to experiment with different caps.

I do notice my caps(ECQ-V1H104JL) on the digikey website have no AC voltage rating, is this an issue with audio being AC? All i want is to protect the opamps without too much loss in overall sound quality.

You don't have something wrong - I have forgotten one thing. Mea maxima culpa...

The 10K summing resistor must be included in the equation. Inverting and and CR in series + feedback is an active filter already. Rin does(!) join the party.
The Formula for the corner frequency stays the same, except that the summing resistor is added instead of the FET input.
1uF for Cin (not 100nF) is the right one to get 16.9Hz low-cut

The given DC voltage rating is OK. 50VDC is about 30VAC, 63VDC = 40VAC, 100VDC = 63VAC.

_________________
Cheers,
Tom.

Some of my projects: TDA2050 Chip Amp, the LM3886 Gainclone Thread and the Szekeres Headamp Thread.

Guess I will order a pile of 1uf's then(Panasonic ECQ-V should do) Any other good values since caps are cheaper to get now then later.

I have been playing with some caps I have. Already noticed the very high pass I was getting with the 0.1uF's so I gave 100 uF a try with my headphones(64 Ohm) and it works without any noticeable issues(shows a 24.87 cut off in the calc you linked me).

I do have a question. The τ(time constant) is a caps charge time?, does this have an effect on music? with the cap not actually getting charged?

I notice it gets into down right strange numbers. So lets say I placed the 100uF caps inline with my mixer(something I can not help but try), I should get about 1s time constant. Does this distort the music, what does this actually mean. All i can find is that it messes with square waves(something music is not).

I took a DPDT switch and setup a test and can not hear anything except at the highest volumes gives a small click when switching or bypassing the capacitor with headphones by using vs bypassing the cap(i am going to guess that its a combo of PC sound not being all that high fi[my mixer has better opamps then the computer ] and cheap headphones, but I like those ones personally.)

A bit low tech(just a basic DPDT), but gets the job done of testing caps. Can be setup on the project board allowing for fast swap testing


As for the board, What I have right now is this. More of the same sockets used for the opamps should allow me to add and remove capacitors or jumpers for sources that do or do not need the filtering, if i can not hear the difference then everything will get input caps.


You may see the mixer has grown to 4 x 6 channel inputs(R L RR RL C LFE). It is just to have it ready for expansion the missing summing resistor does not fit on the image but does look to fit the board. At the top of the board, I have the option to take either the inverted or non inverted output(if I ever wanted inverted for something).

EDIT.

I tried my frankin-capacitor switch on the computer that was putting out 29-30mv and it dropped it way down and it does not make the speakers go POP when plugged in any more. Now I just have to figure out why that computer has a a constant low hum even when connected direct to the speakers(I guess sound kind of ground issue, I have seen it in the past with bad cables too so it may be that as well.).

Yes, tau is the time the cap needs to charge up to 63% of its capacity and one main component of filters. Tau is frequency dependent. When the frequency rises, tau does decrease due to the decrease of the capacitor's Xc (Impedance == 1 / ((2 * pi * f)) * c.
That's why the filter cuts at the calculated frequency. When Xc increases - tau also does increase and the filter dampens the frequency for 3db at the calculated cut-off.
Below the calculated cut-off frequency the dampening rises 20db per decade. Above the corner frequency, tau does decrease (sinks) and the higher frequencies can pass with only a very little time delay in the range of micro seconds.
The delay should not really be audible, because tau has its largest value along the cut-off. When you use caps, they should not be too large, as tau rises with every nF.
Caps with too much capacitance distort the signal as the factor tau also will be higher. They should be in the right range to allow the signal pass through without any flaws.
With AC (music signal) we have other conditions than with DC. With AC we have Xc as impedance and with DC the ESR (serial resistance).
Capacitors are not that bad on the inputs - only when you choose the wrong ones. I never had any problems with input caps and i always like to use them
The sound does not get bad or worse. I've tried many setups before and i can say that a little (only a very very very little) change was audible, but not a bad change.
A little (very very) amount of distortion made the circuits sound a little better. This was caused by the added harmonics (even harmonics are pleasant to our ears) i think and personally i like it.

In fact - the right formula for calculating the Cin is 1 / 2*pi*R*C (== 1 / 2*pi*tau) where R is the 10K summing resistor.
It's the same Formula, but the resistance has changed (Inverting = summing resistor / Non-Inverting = circuit impedance).
Tau is 0.01s along the cut-off and does sink with higher frequencies due to the decrease of Xc. The upper corner frequency is determined by f of the OPA.

Sorry for the mess in the last post again (Apr. 23). I really have mixed up inverting and non inverting. Now - mind becomes clearer. Nightshift has passed and Weekend comes closer

BTW - your mixer is going to be a truly serious project. Good job!

_________________
Cheers,
Tom.

Some of my projects: TDA2050 Chip Amp, the LM3886 Gainclone Thread and the Szekeres Headamp Thread.