Seriously loving this discussion, even though a bit of it is going over my head. 
Thanks for all the suggestions. I'll probably make a decent digikey or mouser order today to sample a few of the recommended opamps and stock up on my discretes. I did after all make my circuit work well, running off of a 12v wall wart instead of USB. For whatever reason, despite the rating, my LME49723 doesn't want to do much of anything at +/- 2.5v. There's still some significant hiss at 12v, but I suspect that has to do with using a cheap and dirty wall wart power supply with minimal filtering. If I had my multimeter here at school with me I could look into things a little further, but alas...
Quick question, when you say "rail to rail" does that refer to the difference in voltage between the + an - on the opamp? So for example an opamp running +/- 2.5v would be 5v rail to rail?
agdr, loving the finding on more opamps I can use with my headphones. Really glad that I seem to have made a good choice.
I actually don't own the headphones yet though... lol. They're due to be released sometime this month. They're the Noontec Zoro II HDs if any of you wanna check em out. Quite a bargain for $99. The original Zoro HDs got great reviews. http://noontec.com/html/en/article_read_278.html
Thanks for all the suggestions. I'll probably make a decent digikey or mouser order today to sample a few of the recommended opamps and stock up on my discretes. I did after all make my circuit work well, running off of a 12v wall wart instead of USB. For whatever reason, despite the rating, my LME49723 doesn't want to do much of anything at +/- 2.5v. There's still some significant hiss at 12v, but I suspect that has to do with using a cheap and dirty wall wart power supply with minimal filtering. If I had my multimeter here at school with me I could look into things a little further, but alas...
Quick question, when you say "rail to rail" does that refer to the difference in voltage between the + an - on the opamp? So for example an opamp running +/- 2.5v would be 5v rail to rail?
agdr, loving the finding on more opamps I can use with my headphones. Really glad that I seem to have made a good choice.
I actually don't own the headphones yet though... lol. They're due to be released sometime this month. They're the Noontec Zoro II HDs if any of you wanna check em out. Quite a bargain for $99. The original Zoro HDs got great reviews. http://noontec.com/html/en/article_read_278.html
The "rails" are just the power supply voltages measured with respect to wherever "ground" is at. In your case the rails are 5V and ground. On a supply with a "real" ground and +12V and -12V outputs the rails are +12 and -12.
The output of op amp usually can't make it all the way to the power supply in either direction. If you supply was +/-12V then a typical thing would be for the op amp to only make it to +/-10V or so on the output. A few op amps can go nearly all the way up to the power supply on the output, within a few milivolts, so those are called "rail to rail" op amps, just referring to how far the output can swing.
There is a similar issue on the input where some op amps don't like having either the + or - input pins taken all the way to the positive or negative rail. They can lock up and do other bad things. So in saying "rail to rail" you even have to specify input or output. That AD8656 op amp is one of the fairly rare ones which can go rail-to-rail on both the inputs and outputs. Rail-to-rail op amps are especially useful for low voltage circuits so none of the potential output voltage swing gets lost.
I've had some fun with your thread here since I've been interested in that "inverting attenuator" circuit with the pot in the feedback loop from a different thread recently. I'm probably going to build one up just to mess with it.
On your amp running on the 12Vdc wall wart, try shorting your input wires together and see if the hiss goes away. That grounds the input so that it is not picking up any stray noise. A more typical thing to get from the wall wart is 60 cycle hum rather than hiss.
Last edited:
It just hit me that both of those problems can be caused by a bad op-amp. Yours may have been damaged when you were getting your wiring straightened out.
Just tried swapping around all three of the LME49723s I ordered, and all still have the slight hiss and whine. Might just be picking up noise due to my sloppy wiring on my solderless breadboard. As I nudge insulated jumpers with my finger I can hear the pitch of the whine change. Still doesn't explain the horrible behavior at 5 volts though.. hmm. Only thing I can think is that I'm not quite getting 5v from my computer's USB port, but like I said I don't have the multimeter to test that. What self-respecting hobbyist doesn't bring his multimeter to college with him?!
I'm gonna place an order on mouser and try building the amplifier you've been talking about, with that small assortment of opamps. Looks quite interesting to me as well, and a good place to go next!
I'm gonna place an order on mouser and try building the amplifier you've been talking about, with that small assortment of opamps. Looks quite interesting to me as well, and a good place to go next!
Update: Switched to a generic 5v cellphone charger wall adapter as my power supply and things improved a lot over my computer's USB port. Makes me think I was right about not getting a full 5v from my computer. Still not what I would call good sounding at all, but sounds like music! Maybe I'm still too close (or too far just below) the voltage threshold for it to behave well. That AD8656 and the NJM4556AM are looking much more interesting now...
000940's comment about not taking the simulation results too seriously is right on the money. Especially at the extremes of any parameters - power supply voltage in this case. Sounds like that LME49723's output voltage isn't going to get as close to the power rails as the sim results suggest. So instead of getting 0.53Vrms you may only be getting 0.1Vrms or worse.
I would expect better results from chips that are made for 5V, like that AD8656 and the OPA2835 that 00940 found. Both those chips are rated for a maximum of 5.5V so definitely designed for 5V usage. Both chips have minimums around 2.5V - 2.7V so that covers some range of fluctuation on the USB 5V bus.
The NJM4556A might not work well either. Even though the datasheet says +/-2V on the low end of the power supply range it didn't simulate well at all on 5V. But that is another chip that isn't really designed for 5V, it goes all the way to +/-15Vdc.
I'm going to send the PC board below out for fabrication and see what happens.
I've sized it to fit a Hammond 1455C801 case, the same one AMB uses for his mini^3 headamp. I tried to fit things into the smaller 1455D801 but the jacks wind up too close to the pot. Turns out the OPA2835 and AD8656 have exactly the same pinout in exactly the same SOIC8 package, so either should work. If the board works I'll post the Gerbers (for making PC boards).Attachments
I was waiting for sgrossklass or 00940 to reply.
They've been posting a lot longer than me and are more current on this stuff. The whole concept of "ground" in electronic things is interesting. It is really just the one single point - anywhere in a circuit - that you chose to make your voltage readings with respect to. In other words, it is the place you put your negative DMM lead.
So with that definition a circuit can actually have several different "grounds", if for some reason you find a reason to make readings with respect to several different points.
For the purposes of the discussion here "real" ground is the one that carries over from some other piece of equipment you are connected to. For a circuit powered by a USB connector it would be the ground the comes along through the USB connector and is connected back in the PC. For a circuit with its own power supply the "real" ground is usually the one used as ground by the power supply circuit.
But what we are doing is making a second point that will also be used as ground, and to give it a different name call it "virtual ground". The thing that is different about our virtual ground is it has a voltage half way between the 5V coming in from the USB connector and that "real" ground in the USB connector, namely 2.5V measured with respect to real ground. Note that since we've defined a second ground somewhere it is now necessary to specify with each and every voltage reading which of those two grounds the measurement is to be taken with respect to (where you put the meter negative lead).
How that 2.5V virtual ground point is created is just two 4.7K resistor in series, a resistive divider. The voltage at the center point will be half what is applied, 2.5V in this case, and will more or less stay at that voltage as long as any current extracted from that point is 1/10 or less of what is going through the two series resistors.
As for why a virtual ground is needed, take a look at a typical sine wave. It swings as much positive as negative about the zero point.
http://en.wikipedia.org/wiki/Sine_wave
In a piece of equipment that has positive and negative power supplies, with "real" ground in the middle, the output transistors can swing as far in the positive direction as the negative since "real" ground is in the middle between two equal voltages. But for a single power supply, with "real ground" simply at the lowest voltage level available, now we have a problem. The positive half of the sine wave could swing up from ground, but the negative half would have nowhere to go since you are already at the lowest voltage in the circuit. By creating a virtual ground at 1/2 the power supply voltage the output transistors in a chip once again can swing as far positive (+2.5V) as they can negative (-2.5V).
So here is a quiz question.
What voltage would you get if you measured the voltage between the two series 4.7K resistors with respect to virtual ground rather than real ground?I'll leave your second question for sgrossklass or someone else.
Last edited:
matdogg - something else interesting, after some reading. I had forgotten about the 100mA/500mA high and low current modes on the USB bus, Turns out that until anything plugged into the USB port has had a discussion with the host CPU (enumeration) only 100mA will likely be supplied. You should still be OK at 100mA, since your headphones would draw 25mA peak, times two channels, then a few mA more for chip quiescent current.
But for higher current draws, like 50mA per channel with that AD8656 chip up near the 120dB peak SPL, that 100mA/500mA issue has to be dealt with. I'll take a guess a chip exists out there somewhere that can deal with the 500mA high current "handshake" data exchange with the PC, without having to put a microcontroller on the amp.
Also here is an diagram of a USB powerline noise suppressor in figure 2.5 of the PDF below. Just a C-L-C filter with a chip ferrite bead in for the inductor. Figure 2.4 in there shows an inrush current limiter to use with caps bigger than 10uF across the USB supply line. The revised CMOY circuit has that stuff added, along with 00940's suggestion of the 470uF rail to ground cap after the inrush limiter. The blue line in the plot is the rail voltage after the current limiter.
But for higher current draws, like 50mA per channel with that AD8656 chip up near the 120dB peak SPL, that 100mA/500mA issue has to be dealt with. I'll take a guess a chip exists out there somewhere that can deal with the 500mA high current "handshake" data exchange with the PC, without having to put a microcontroller on the amp.
Also here is an diagram of a USB powerline noise suppressor in figure 2.5 of the PDF below. Just a C-L-C filter with a chip ferrite bead in for the inductor. Figure 2.4 in there shows an inrush current limiter to use with caps bigger than 10uF across the USB supply line. The revised CMOY circuit has that stuff added, along with 00940's suggestion of the 470uF rail to ground cap after the inrush limiter. The blue line in the plot is the rail voltage after the current limiter.
Attachments
Last edited:
I was waiting for sgrossklass or 00940 to reply.
The whole concept of "ground" in electronic things is interesting. It is really just the one single point - anywhere in a circuit - that you chose to make your voltage readings with respect to. In other words, it is the place you put your negative DMM lead.
So with that definition a circuit can actually have several different "grounds", if for some reason you find a reason to make readings with respect to several different points.
For the purposes of the discussion here "real" ground is the one that carries over from some other piece of equipment you are connected to. For a circuit powered by a USB connector it would be the ground the comes along through the USB connector and is connected back in the PC. For a circuit with its own power supply the "real" ground is usually the one used as ground by the power supply circuit.
But what we are doing is making a second point that will also be used as ground, and to give it a different name call it "virtual ground". The thing that is different about our virtual ground is it has a voltage half way between the 5V coming in from the USB connector and that "real" ground in the USB connector, namely 2.5V measured with respect to real ground. Note that since we've defined a second ground somewhere it is now necessary to specify with each and every voltage reading which of those two grounds the measurement is to be taken with respect to (where you put the meter negative lead).
How that 2.5V virtual ground point is created is just two 4.7K resistor in series, a resistive divider. The voltage at the center point will be half what is applied, 2.5V in this case, and will more or less stay at that voltage as long as any current extracted from that point is 1/10 or less of what is going through the two series resistors.
As for why a virtual ground is needed, take a look at a typical sine wave. It swings as much positive as negative about the zero point.
Sine wave - Wikipedia, the free encyclopedia
In a piece of equipment that has positive and negative power supplies, with "real" ground in the middle, the output transistors can swing as far in the positive direction as the negative since "real" ground is in the middle between two equal voltages. But for a single power supply, with "real ground" simply at the lowest voltage level available, now we have a problem. The positive half of the sine wave could swing up from ground, but the negative half would have nowhere to go since you are already at the lowest voltage in the circuit. By creating a virtual ground at 1/2 the power supply voltage the output transistors in a chip once again can swing as far positive (+2.5V) as they can negative (-2.5V).
So here is a quiz question.
I'll leave your second question for sgrossklass or someone else.
Thank you for the explanation.
Before reading your reply, I had it in my mind that real ground was the ground from the power supply and virtual ground was the audio signal ground.
So I was right about one.
As for your quiz question, I don't have a clue as I've not familiar with voltage dividers and such.
If you remember, I'm the one who posted the circuit with the volume pot in the feedback loop.
Another member stated it was a bad idea, so I went back to keeping the feedback loop fixed and using the pot to control the audio signal going to the first stage.
I see that you like it well enough to have some boards made up and actually try it.
Just curious...how do the CMOS op amps, like you've been working with, compare to FET op amps like the OPA2134 or AD8620?
I see that they operate on much lower voltages, but are there any advantages or disadvantages to using them over the FET devices above?
Is sound quality about the same or better?
Thanks...
That's theory... in practice, almost no usb port will actually limit the current delivered to 100ma. Otherwise all those usb gadgets wouldn't work.
The only exceptions might be tablets and smartphones, which attempt to preserve energy.
Hey that is good to know! Thanks.
Explains why I haven't been able to find an IC to deal with it yet.Knowing about potential current limiting circuitry on the upstream side of the USB port, and/or circuitry to switch the port off altogether, I'm not expecting good things from the output impedance in the audio band looking back into that port. That 470uF or so bypass cap you suggested across the power lines is probably going to turn out to be fairly important.
After I wrote that I almost put TLDR: at the bottom, lol.
The whole thing can be summed up to say the virtual ground, in between the two 4.7K resistors, biases the op amp at 1/2 the power supply voltage so the signal coming out can swing equally in the positive and negative voltage directions.The answer to the quiz is zero volts. That point between the two 4.7K resistors IS virtual ground. so measuring there with respect to that same point would just yield zero volts. Typical test trick question!
Yeah your original post got me quite curious about the inverting attenuator when it turned out that was actually a stable circuit. I never thought it was! I think it was 00940 who had the objection on the pot in the loop in your thread. But he also did post the solutions to mitigate the problems, that DC blocking cap in series with the pot (to keep the DC out of the wiper) and the 200K in parallel with the pot to keep the loop closed when the wiper eventually dies years down the road. He was quite right about both of those being potential problems.
I've actually never used a CMOS op amp for audio before. That is another reason I'm kind of interested in this one. NwAvGuy gave them the thumbs down in his op amp review article here:
NwAvGuy: Op Amp Measurements
Scroll down to the "Don't Bother" heading. But then he qualifies that with "Unless you have a specific reason (such as a strict low power/low voltage requirement)", which is exactly the case here with the USB 5V. Plus this particular AD8656 chip does seem to be designed for audio. The voltage noise and THD specs they show are right up there.
Here is the answer to your second question about the cap to ground in the feedback loop. It is the difference between DC and AC gain, with DC defined just as a signal at 0 Hz. At DC, 0 Hz, that cap to ground is an open circuit. So you can re-draw the feedback loop as just the resistor from output to inverting input, which is a unity gain voltage follower. Whatever DC offset the chip has internally will only get multiplied by 1. But at AC frequencies > 0 that cap now has some low(er) impedance and the resistor to ground in the loop is back in the circuit, increasing gain past 1 for AC signals.
The cap to ground is just a neat trick to keep DC gain at 1 while AC gain is something higher. But the downside is it puts a fairly large value (electrolytic) cap in the signal path.
I spotted another trick in the reading I think I'm going to add to the board, a common mode choke right after the USB port power pin. Apparently the single ferrite bead chip is useful for reducing differential mode noise, but does nothing for common mode noise, which I expect will be high with PC clock harmonics getting into everything.
It turns out the AD8656 is good for a whopping 220mA per channel at 5V. I was looking at the wrong table, the 2.7V power supply table, which shows the 75mA max output. Good grief, if it can really do a 1.6Vpeak swing and up to 220mA out that would be useful for several headphones. With the low voltages involved power dissipation in the chip won't be a problem. If that attenuator function on the low end of the pot range works as predicted the amp would even work to cut the level of high output sources, for phones that just need current buffering.
Attachments
Last edited:
But does it degrade the sound like a cap would in series with the +input or in series with the output of the op amp?
Thanks...
Essentially yes, since it winds up in the signal path just like an input or output coupling cap. But if you use a low ESR cap it may be hard to tell any difference between having a cap involved or not. Note: low ESR (equivalent series resistance, the internal resistance of the electrolytic) is not the same as caps sold as "audiophile" caps - some of those have fairly high ESR. Just a regular electrolytic but one sold as low ESR, like 50mR (0.050 ohms). The organic polymer electrolytics I'm using this this inverting CMOY all have extremely low ESRs. Those caps are relatively cheap these days.
Last edited:
For this inverting CMOY I'm using Mouser #667-6SEPC470MX for the 470uF 6.3V output caps. Only $1 each, not bad as these things go. They are listed as ESR = 8mR (0.008 ohms) at 100KHz. Looks like they are something new, a Panasonic "SEPC" series of their Oscon line.
Film caps are preferable to electrolytic for audio signal path use but are only available in smaller capacitances. I used 3.3uF film for the input coupling caps and the feedback caps in the inverting CMOY (Mouser #505-MKS23.3/50/10). But since the output coupling caps needed over 1000uF due to the low impedance they are driving (forms a resistive divider with that headphone load) there was no other choice than electrolytic.
Last edited:
To finally get back to this one:
I'd go a bit higher than 10µ in the old tradition of oversizing electrolytics by a factor of 10, but it should work.
As far as choosing output coupling caps goes, there probably are multiple ways to skin a cat, as usual. Just keep in mind that it typically is not possible (or economic) to make capacitor impedance negligible throughout the bass region - unless you feel like using 2200µ caps for 32 ohm loads. A very low-ESR part may do well throughout most of the audible range but might not down there. Case in point: Tantalums. Bipolars or "DIY bipolars" may prove useful, at least they have in capacitor distortion measurements.
At least it'll make sure the offset is not being amplified further, as for f-->0 gain drops to unity as capacitor impedance goes through the roof, with only the 2k shown in the feedback remaining active.
I'd go a bit higher than 10µ in the old tradition of oversizing electrolytics by a factor of 10, but it should work.
As far as choosing output coupling caps goes, there probably are multiple ways to skin a cat, as usual. Just keep in mind that it typically is not possible (or economic) to make capacitor impedance negligible throughout the bass region - unless you feel like using 2200µ caps for 32 ohm loads. A very low-ESR part may do well throughout most of the audible range but might not down there. Case in point: Tantalums. Bipolars or "DIY bipolars" may prove useful, at least they have in capacitor distortion measurements.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.