One of the flaws in the human psyche is that we all have a tendency to believe that we are better than everyone else, more capable, more lucky. (This is the only reason why people buy lottery tickets, for example, since statistically it's quite obvious that a person will only lose money with every ticket he/she buys.)
Education, and self-restraint, is the only way to avoid falling into this very common, and entirely unconscious belief that "I'm more special than everyone else".
In this case, we have the results of many decades of carefully conducted experiments on human hearing, encompassing many different researchers, many different research methodologies, and tens of thousands of subjects.
And the result is that 1% THD is just barely detectable to a significant percentage of people, and 0.1% is undetectable.
Plenty of people who post on diyAudio think they can hear 0.0003% distortion or 100 volts/microsecond slew rates. There is no evidence to support it, only that unconscious human tendency to self-delusion. It's no different than people who think they can tell the future, or speak to ghosts, or bend teaspoons with the power of their minds. Or, for that matter, buy lottery tickets!
The TL084 datasheet specifies 0.003% THD at unity voltage gain. THD increases roughly linearly with open-loop gain in this sort of circuit, we could configure our TL084 for a voltage gain of 300 times, and still expect THD levels below audibility. More realistically, we could configure a gain of x30, and be confident that the THD is at least ten times too little to hear.
You misunderstood. The slew rate required to output a sinewave depends on two things, frequency, and amplitude. I wasn't referring to whether the output gets within millivolts of the output rails (as some opamps can), but rather to the fact that the TL084 has more than six times the slew rate needed to output it's maximum possible unclipped signal voltage. It's really irrelevant whether that maximum is 30 Vpp, or 24 Vpp - these are semiconductor devices we're talking about, everything has substantial tolerances on it.
If you're hair-splitting, well, you forgot the source impedances of the signal sources connected to the inputs.
But if you know enough to calculate the Thevin impedance of a resistive network, you also know that pots come with large tolerances in resistance. And also that it takes about a 25% change in resistance to produce a 1 dB change in thermal noise.
Because of all this, saying that the noise voltage is equivalent to a 27.4 k resistor rather than a 25k one is like trying to compute, to three decimal places, the number of cats that will fit into one cubic metre.
It's your money, and your decision, of course you can choose any opamp you like for any project of yours, for any reason whatsoever. Including that "OPA" is a Greek word, while "TL0" isn't a word in any language.
But that isn't the same thing as telling someone else - who's relying on the advice he gets on this forum - that a TL084 is "dirty, noisy, and slow", and an OPA1654 will be vastly better in his circuit. Neither of those claims is accurate.
-Gnobuddy
One nit to pick
<mode="olde pharte">
</mode>
However, that's a nit. And I digress.
My practical experience with misbehaving op-amp circuits revolve around three related issues. Bob Pease would add a few more.
Firstly, don't ask the op-amp to do things the designers didn't intend. (Unless you're Bob Pease). Especially clipping it. Corollary: non-DC gain and output current are finite.
Secondly, clean power and ground references. If your input is NOT referenced to the same ground as the op-amp's V+, V- and the other input you're robbing the op amp of any chance to perform. Corollary: CMRR is a theoretical starting point.
Thirdly, and really just looking at the first two through a different lens, keep big and small signals apart. Almost everything couples somehow
Corollary: the physical layout matters. A lot.
There's a lovely concept called "noise budget" which you have to play off against your dynamic range at any point - some where there's a rather good discussion of how this works in the context of a typical recording chain (by Rupert Neve, I think).
<mode="olde pharte">
....is about as much use as GDP as a measure of merit
</mode>
However, that's a nit. And I digress.
My practical experience with misbehaving op-amp circuits revolve around three related issues. Bob Pease would add a few more.
Firstly, don't ask the op-amp to do things the designers didn't intend. (Unless you're Bob Pease). Especially clipping it. Corollary: non-DC gain and output current are finite.
Secondly, clean power and ground references. If your input is NOT referenced to the same ground as the op-amp's V+, V- and the other input you're robbing the op amp of any chance to perform. Corollary: CMRR is a theoretical starting point.
Thirdly, and really just looking at the first two through a different lens, keep big and small signals apart. Almost everything couples somehow
Corollary: the physical layout matters. A lot.
There's a lovely concept called "noise budget" which you have to play off against your dynamic range at any point - some where there's a rather good discussion of how this works in the context of a typical recording chain (by Rupert Neve, I think).
Last edited:
I don't know where you determined that...the paper I read said
"Distortion energy was detectable at an average level of 94 db below the level of the signal in capacitors and an average level of 86 db below the signal level in op amps within a 48 KHz bandwidth"
-86 db would be ~0.005% THD. That is FAR below what you are quoting.
It's NOT "irrelevant". It's called headroom and a rail-to-rail opamp like the OPA1654 will give ~25 db of headroom, about 2 db more than a TL084.
Indeed!!If you're hair-splitting, well, saying that the noise voltage is equivalent to a 27.4 k resistor rather than a 25k one is like trying to compute, to three decimal places, the number of cats that will fit into one cubic metre.
Well, if you are building a mixer such as the OP proposed, the cost of the rest of the project will swamp any differences in opamp expense---in this case, about $2 between a TL084 and a OPA1654. For ANYBODY's money, I still say that the OPA1654 is a bargain; it provides state-of-the-art performance for little cost. By your arguments, ANY op amp will do just fine, as they are ALL well below 0.1% (even the terrible RC4558). I just don't think that's a good approach when trying to build the best that you possibly can.
First of all, thanks for all the reasoning around different op-amps and noise levels, taught me a lot. Then on to the next problem..
Ok, I should have read this more carefully.... So in my retarded mind, I thought that I should use the ground that comes out of my dual supply for the op amp grounding, since I saw this as an virtual ground which I thought the op-amp would need for reference. And for the audio grounds, I'd use the "real ground". This did not work very well, now I have to wait for a new PSU... My glass fuses blows as soon as I turn on the power, even with the PSU "floating". Should the "ground" from the single to dual supply be left unconnected simply, since this would really be +12V in reference to "real" ground? My active x/o has the ground connected and it seems to be working fine.
My setup is a 24V single polarity SMPS PSU, Protective Earth separated from V- (is that called isolated? I get 0V when measuring from V to PE). This is feeding three things; my TDA7498 amp and two single-to-dual +/-12V supplies to feed an active crossover and the mixer.
I'm having trouble grasping the concept obviously so I'd love if someone could guide me.
I've got eight 3,5 mm jacks mounted to a metal panel, it's really hard to find mountable insulated ones at a reasonable price, so unfortunately there will probably be some noise from here.. Not sure what I can do about it, so I guess I'll have to live with it. This panel *should* be grounded by PE, but since the PE is separated from V-, I get confused. Should V- from the PSU be connected to PE? I'm asking about this in another thread as well, so I probably shouldn't go on about it here..
Ok, I should have read this more carefully.... So in my retarded mind, I thought that I should use the ground that comes out of my dual supply for the op amp grounding, since I saw this as an virtual ground which I thought the op-amp would need for reference. And for the audio grounds, I'd use the "real ground". This did not work very well, now I have to wait for a new PSU... My glass fuses blows as soon as I turn on the power, even with the PSU "floating". Should the "ground" from the single to dual supply be left unconnected simply, since this would really be +12V in reference to "real" ground? My active x/o has the ground connected and it seems to be working fine.
My setup is a 24V single polarity SMPS PSU, Protective Earth separated from V- (is that called isolated? I get 0V when measuring from V to PE). This is feeding three things; my TDA7498 amp and two single-to-dual +/-12V supplies to feed an active crossover and the mixer.
I'm having trouble grasping the concept obviously so I'd love if someone could guide me.
I've got eight 3,5 mm jacks mounted to a metal panel, it's really hard to find mountable insulated ones at a reasonable price, so unfortunately there will probably be some noise from here.. Not sure what I can do about it, so I guess I'll have to live with it. This panel *should* be grounded by PE, but since the PE is separated from V-, I get confused. Should V- from the PSU be connected to PE? I'm asking about this in another thread as well, so I probably shouldn't go on about it here..
When I think about it, I guess the bypass caps are the only things that should be connected to the dual supply ground, and the rest to audio/common ground? the op amp compares the two inputs right? As I connected it, the "ground" was set to +12V... So, do you think I also busted my op-amp then?
What you just described is an "internal grounding problem" with your mixer. It is a common problem encountered when trying to connect "home audio" and "pro audio" equipment together. In a piece of equipment, there are three distinct grounds: power supply, signal and chassis/case. Each has a function and all 3 interact.
When using (or designing) battery powered equipment, it is OK to float the power supply as you describe, as there is no return path between the battery and earth for current to follow. When connected to what have you, the equipment will naturally float the audio ground to whatever the external components supply. Nearly all "battery eliminator" power supplies float their leads/plugs (that connect to powered devices) by not connecting the transformer secondary to a reference. The case of the equipment is not directly grounded to earth. The audio ground is tied (directly or indirectly) to the case. This is how most "home audio" equipment is designed.
Nearly all "non-portable" equipment isolates the AC supply leads by using the power transformer for power line isolation. If the equipment is grounded, the power plug ground wire goes right to the case at a single point. When designing the equipment, all other circuits, inputs, outputs, etc. are designed around this single "global" earth ground. There are several rigid tests that are done on input and output terminals on the case to address safety, interference, and noise issues. This method of grounding is how "pro audio" equipment designers do it.
With your mixer, it would be much simpler to rig a power supply that follows one of these two methods. Either way, your derived audio supply ground should be tied to the case. If your manufactured power supply has the "V-" lead tied to anything (typically one of the input power leads or its own case), it won't work without making a compromise.
Last edited:
Thanks jeff5may, what would I do without this forum?
I've decided to go with a proper transformer and diode bridge, etc. This way I'll get all grounds to line up. Even if I could get it to work, I think I'd have problems if i connected something with a "normal" ground potential to my mixer right?
If I'd looked at it more closely from the beginning, it would have saved me some of this BS I just thought the op-amps would amplify the signal, using the power from the dual PSU. Didn't understand the proper function at first.. But hey, we call it "learning-money".
I've decided to go with a proper transformer and diode bridge, etc. This way I'll get all grounds to line up. Even if I could get it to work, I think I'd have problems if i connected something with a "normal" ground potential to my mixer right?
If I'd looked at it more closely from the beginning, it would have saved me some of this BS
I just thought the op-amps would amplify the signal, using the power from the dual PSU. Didn't understand the proper function at first.. But hey, we call it "learning-money".Hey guys, got it all working now and I'm pretty happy with it all! Almost no noise at all, which is above expectations considering there's a lot of wires and grounds running all over the place
But, I have a couple of "problems":
1) I guess it's just the way this simple mixer works, but if I pan a channel, the summed output to the speaker gets lower. Makes it a bit hard to know the actual level of the input, which gets recorded on line out. Not much to do though I think..
2) The output from the mixer is pretty low, I have to crank the volume up quite a bit and it's not really loud enough even on max level.. The level on the line out is OK though. I guess I could change the resistors over the op-amps to increase the gain? Or would I need a preamp between the mixer and amp?
Any thoughts? I'll upload a couple of pictures in a bit so you can see the monster
But, I have a couple of "problems":
1) I guess it's just the way this simple mixer works, but if I pan a channel, the summed output to the speaker gets lower. Makes it a bit hard to know the actual level of the input, which gets recorded on line out. Not much to do though I think..
2) The output from the mixer is pretty low, I have to crank the volume up quite a bit and it's not really loud enough even on max level.. The level on the line out is OK though. I guess I could change the resistors over the op-amps to increase the gain? Or would I need a preamp between the mixer and amp?
Any thoughts? I'll upload a couple of pictures in a bit so you can see the monster
So, it's not quite finished yet. Still waiting on some knobs and cables.
The bottom panel holds a usb midi host with in/out, Midi thru 1 in 4 out, clock sync by 3.5mm plugs, line out with level and level for the speaker and standby swith.
Right panel holds 8 channel mixer with panning, the top of that panel is the breakout for an Axoloti synth (stereo in/out, midi in/out, usb host/in) which will be housed in the lid of my flight case. Also 3 outputs which comes from three Teenage Engineering pocket operators, that will also be mounted in the lid.
Here's the rats nest.. Not the most astounding craftmanship, but it's functional
And here's where it'll sit. In the process of designing the lid now, where the axoloti will be sitting. I will have something like 10-15 potentiometers and some buttons to control it.
I'm also working on a mechanism which will allow me to raise the case to a 30 degree angle. Not the best acoustics when the speaker is pointing to the ceiling
The bottom panel holds a usb midi host with in/out, Midi thru 1 in 4 out, clock sync by 3.5mm plugs, line out with level and level for the speaker and standby swith.
Right panel holds 8 channel mixer with panning, the top of that panel is the breakout for an Axoloti synth (stereo in/out, midi in/out, usb host/in) which will be housed in the lid of my flight case. Also 3 outputs which comes from three Teenage Engineering pocket operators, that will also be mounted in the lid.
Here's the rats nest.. Not the most astounding craftmanship, but it's functional
And here's where it'll sit. In the process of designing the lid now, where the axoloti will be sitting. I will have something like 10-15 potentiometers and some buttons to control it.
I'm also working on a mechanism which will allow me to raise the case to a 30 degree angle. Not the best acoustics when the speaker is pointing to the ceiling
Also, I'm losing a lot of low-end. when I bypassed the mixer and volume control, it sounds normal, so it's not my active X/O I'm happy to say. I have a strange feeling it might be my volume control acting up though, I've put together the "fake log-pot" from ESP. Gonna try to bypass that next and see what happens. The reason I believe that, is because when I recorded the mix from Line out, I didn't notice the drop of low end.
Nice build, you put a lot of work into that!
I went back to the first page and looked at your schematic. And I saw the problem almost immediately - your input coupling caps are only 0.1 uF. If volume is at max and the corresponding "pan" pot is at one end, input resistance is less than 20 kilo ohms.
If you put 0.1 uF and 20k into the RC highpass filter formula, you get a corner frequency of about 80 Hz - frequency response is down about 3 dB at 80 Hz, and everything below that is rolled off at 6 dB/octave.
80 Hz bass cutoff is fine for guitar, and probably fine for almost all vocals. But definitely not okay for bass guitar, kick drums, synth bass, and so on.
This is assuming "real" 100k log volume pots. If you're using 100 k linear pots with a 22k taper resistor in parallel (just a guess; RG Keehn suggests one-fifth of the pots value), the pot has an effective resistance around 20k at full volume. If the pan pot is also at either extreme, the input resistance of that channel is now only 10 kilo ohms. Now the bass cut-off frequency is 160 Hz - everything below that is attenuated. This is definitely inadequate bass response, and will really thin-out the sound badly.
The simplest cure is to increase those 0.1 uF caps to 1 uF caps, which will push the corner frequency down to around 16 Hz, worst case, even if you're using 22k taper resistors with 100k linear pots.
I don't think increasing the volume pots will solve the issue, because the network around the pan-pots is just as big a problem.
Caps are cheaper than pots anyway, so replacing the caps is probably the way to go. (Don't worry about using expensive cap types, the cheapest ceramic 1 uF caps are perfectly fine for the job. The tiny amounts of distortion they might produce are below the audibility threshold, so they sound exactly the same as the overpriced ones.)
-Gnobuddy
I've not had the time to test the circuit more after my last post, being a dad really eats up your time for hobbies!
Great find! Seems like a strange mistake in the circuit to make, especially since it's a page committed to synths.. hmm, not looking forward to resolder all that. I'll change one channel and listen to the difference. I think I have 1uF caps of good quality lying around, otherwise I also have some ceramic ones. Thanks for the tip gnobuddy, I'll let you know how it works out tomorrow!
Great find! Seems like a strange mistake in the circuit to make, especially since it's a page committed to synths.. hmm, not looking forward to resolder all that. I'll change one channel and listen to the difference. I think I have 1uF caps of good quality lying around, otherwise I also have some ceramic ones. Thanks for the tip gnobuddy, I'll let you know how it works out tomorrow!
I can only imagine. I hope it all feels worthwhile in the end, after all, hobbies don't love you back!
I agree, this is a pretty basic mistake, and should not have happened.
Perhaps the designer never considered what happens with controls set to extreme positions, and so calculated a much higher input impedance. But that too is a pretty major oversight. There is a reason why passive mixers aren't more popular, and it has to do, in part, with managing input impedance.
And finally, there is the question of why the designer never heard a problem with his prototype. All he had to do was use his ears, as you did, and the problem should have become obvious.
Good plan!
I wish I had noticed the problem as soon as you started the thread. That would have saved you some wasted time, effort, and money. Unfortunately, I didn't catch it at first glance.
-Gnobuddy
Hey again!
So I've done some extensive testing now.
-I have bypassed the mixer completely, using only a pot to adjust the volume.
This yields a much higher gain, so I'm clearly losing a lot over the mixer.
-I didn't have 1uF caps as I thought, so I took a .1uF and used it together with the pot to see if I lost low end.
I couldn't hear any difference really when using the cap or bypassing it. I used a function generator and set it to 50-60Hz and the speaker gave the same sound level with or without the cap.
-Bypassed mixer and used my self-tapered vol pot to control the level.
Lower max level, but I didn't mind the profile of the linear pot I tried before, so I might skip the tapering resistor.
-Listening again made me realize maybe I don't lose any low end after all, maybe a bit when the level is turned way up as suggested. I think the difference in level w/ or w/o the mixer fooled me a bit.
-I used my phone to play music I know how it is supposed to sound, but the output of my phone apparently is ****. When I connected a synth, the level was quite loud. But since I have a bunch of stuff with varying output levels/impedance, I need to increase the gain so that they all can be used together.
So, resoldering those caps are going to be a real pain in the ***. I think it might have to do if I can increase the gain in the circuit, so that I don't have to turn up the levels as high. Would it be enough to change the 39kOhm resistors to increase the gain? Or is this a bad solution in any way? I guess the other solution would be a pre-amp, but it feels like this should be roughly the same as increasing the gain in the circuit?
So I've done some extensive testing now.
-I have bypassed the mixer completely, using only a pot to adjust the volume.
This yields a much higher gain, so I'm clearly losing a lot over the mixer.
-I didn't have 1uF caps as I thought, so I took a .1uF and used it together with the pot to see if I lost low end.
I couldn't hear any difference really when using the cap or bypassing it. I used a function generator and set it to 50-60Hz and the speaker gave the same sound level with or without the cap.
-Bypassed mixer and used my self-tapered vol pot to control the level.
Lower max level, but I didn't mind the profile of the linear pot I tried before, so I might skip the tapering resistor.
-Listening again made me realize maybe I don't lose any low end after all, maybe a bit when the level is turned way up as suggested. I think the difference in level w/ or w/o the mixer fooled me a bit.
-I used my phone to play music I know how it is supposed to sound, but the output of my phone apparently is ****. When I connected a synth, the level was quite loud. But since I have a bunch of stuff with varying output levels/impedance, I need to increase the gain so that they all can be used together.
So, resoldering those caps are going to be a real pain in the ***. I think it might have to do if I can increase the gain in the circuit, so that I don't have to turn up the levels as high. Would it be enough to change the 39kOhm resistors to increase the gain? Or is this a bad solution in any way? I guess the other solution would be a pre-amp, but it feels like this should be roughly the same as increasing the gain in the circuit?
Excellent! Yes the caps are too low in value. However, the white paper I read said:
Distortion energy was detectable at an average level of 94 db below the level of the signal in capacitors within a 48 KHz bandwidth
I would change the 20 Ks to 10K and put 2.2uF polypropylene caps in---they are only a buck apiece, anyway. If you were building 5,000 of these----sure, cost would be a BIG issue. But for $4, the polypros are the way to go, methinks.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.