DC offset in my preamplifier design?

Discussion in 'Analog & Mixed-Signal Design' started by Sulkyoptimism, Oct 7, 2018 at 8:45 PM.

  1. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Hello there,
    I signed up just to ask this as this seems like a pretty reactive group and knowledgable at that.

    I'm designing and simulating a preamplifier that has multiple EQ's and different routing options, my issue is that im not entirely sure how DC offset is generated. Usually just following guidlines on Coupling Caps, etc would be okay but when all of these caps are added the signal gets very badly diminished in bandwidth. I've read about buffer amps but not entirely sure how that concept would work either. There are just too many stages between all of the EQ segments, specifically a cascaded series of parametric circuits to have caps inbetween each, can i just remove them? Does it matter when its so internal etc?

    If possible can someone clearly explain either buffer amplifier circuits and how they remove DC offset or where DC offset is generated and how i could measure it either through simulation or testing?

    I'd be very grateful, thanks to any that can help!
     
  2. MrChips

    Moderator

    Oct 2, 2009
    16,817
    5,160
    Welcome to AAC!

    A DC voltage is created when a DC current is allowed to flow through any significant resistance.
    Generally speaking, DC is blocked from one audio stage to the next using a series DC blocking capacitor. This creates a high-pass filter which allows AC signals to pass while preventing the flow of low frequency signals. For audio frequencies from 20Hz to 20kHz, you want the filter to pass frequencies greater than 20Hz and start attenuating frequencies below 20Hz.

    Sometimes this calls for a DC blocking capacitor with values in the 10 to 100μF range. Capacitors can build up a charge if the resistance required to bleed off the DC voltage is too large. This can happen if the input resistance of the circuit is too high, such as with preamp circuits. To prevent this from happening, you need to put a bleeder resistor from the output lead of the capacitor to ground.

    You will have to show actual circuits.The devil is in the details.
     
    Sulkyoptimism likes this.
  3. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Thanks for the welcome MrChips,
    Glad to be here.

    So where exactly is the DC generated, is it because of the 3 way split power rails being on ground? I know that its Particularly bad over voltage dividers or pots, but how does this come about? Is there an equation? Or a simulation that i can run? I want to avoid having to build it to find the Dc values.

    What you've said confirms my understanding of the siutation so far, I'm looking at some other posts on here for selecting values etc, but I want to know about other solutions and whether they are absolutely needed or i can reduce the number of them in my design.

    I've posted a picture of the area in question, two designs are shown one without the caps and one with the reccomended caps for the design, as you can see one has many many caps causing a compounding of the -3db at 20 hertz effect. Would it be appropriate to change the values until the 20hz reading is indeed at the optimal 3dB? I imagine this would have the same effect just slightly more noise. The issue is im not sure whether i can measure the DC offset voltage, to change the design without literally building it and testing it that way, even then im unsure what to do.

    Any help as always is great,
    Thanks to MrChips for his answer,
    Regards,

    Sulky
     
  4. Veracohr

    Well-Known Member

    Jan 3, 2011
    649
    98
    What amplifier model are you simulating with? And where do the non-inverting inputs and the outputs go that is labeled "Process"?

    If the non-inverting inputs are grounded, as would be typical for an inverting amp, and if the amplifier is a BJT-input model, then the input bias current combined with the 22k resistors can cause offset at the output. See here for an example of the effect and how to cancel it with a resistor at the grounded non-inverting input:

    Screen Shot 2018-10-08 at 5.05.39 PM.png


    Another option is to use a JFET-input opamp like the TL07x, which has negligible input current. Also, you may be able to disregard small amounts of offset (a few mV) passing from one stage to the next, unless you have a stage with large gain.

    But if you really want to use capacitive coupling between every stage, you need to set the cutoff frequency much lower so that as the attenuation stacks up from stage to stage, it has insignificant effect on the frequency range of interest. You can make the input stage limit the bandwidth to your desired range.
     
    Sulkyoptimism likes this.
  5. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Hello there,

    The noninverting and output go to potentiometer/device that fedds into a feedback process that is pretty much a optocoupler stereo controlled parametric eq, there are 5 cascaded into one another, the circuit is quite large and very difficult to get in a photograph, i thought this would be the most relevent part.
    I've tried altering the caps, but unsure what range i should be looking for 0Hz to be at? I assume thats where the DC voltage is at?
    You've lost me a little with the amp talk there, my apologies i'm self taught so i'm maybe not the best at keeping up.
    I use LTspice currently.

    Is there any other information i can supply to help with this?

    Thanks for your help i'll no doubt end up looking these options up.
     
  6. Audioguru

    Expert

    Dec 20, 2007
    10,454
    1,151
    In your first circuit you have two capacitors in series and two resistors in series. Why?

    The 10uF value of the coupling capacitors is much too high for audio and they are polarized with simple guessing about correct polarity.
    1uF into 22k passes earthquake frequencies down to and below 0.7Hz. Audio does not go below 20Hz so why not use non-polarized 1uF film capacitors?

    1uF film capacitors are expensive then simply increase the value of all 22k and 2.7k resistors 3 times to 68k and 8.2k, then use inexpensive 330nF (0.33uF) film capacitors.
     
  7. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    I'm not entirely sure, but that was the set up that was shown, Could i remove a pair? I did wonder if it was going to be overkill.
    Sorry am i crazy or will reducing the caps to 1uF not have a significant highpass, removing the bass?

    Also sorry what does earthquake frequencies mean? Seems obvious but in this context im not sure.

    Thanks for the time to help me with this though!
     
  8. Audioguru

    Expert

    Dec 20, 2007
    10,454
    1,151
    You do not need two series coupling capacitors, remove one of them and when 100 ohms is in series with the 22k ohms then the 100 ohms does nothing and can be replaced with a piece of wire.

    You cannot hear frequencies below 20Hz, instead you feel them as earthquake vibrations. Bass sounds are from 20Hz to about 200Hz.
    Most speakers have difficulty producing sounds down to 40Hz and most do not produce vibrations below 20Hz.
    So your coupling capacitors have way too much capacitance and your resistor values are too low.
     
    Sulkyoptimism likes this.
  9. Veracohr

    Well-Known Member

    Jan 3, 2011
    649
    98
    I’m assuming you’re just running simulations for now. How much DC offset does a single stage produce, and what model of opamp is used in the simulation?

    I’m trying to help you find a way of either canceling out the offset or not creating it in the first place, rather than use coupling capacitors.
     
    Sulkyoptimism likes this.
  10. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Ah okay i see what you mean, are you talking about load resistors or the ones connected to the ooutput of the caps, bleed resistors? I've put the caps all the way up to 35uF now and they pass the level i need,now give -3dB at 20Hz, but like you said i want to reduce those values, what would i substitute in? I've included a photo of ESP's State variable parametric, which is what i based mine off of. Please help aha, i feel like im so close.

    Yeah just sims for now, Also the schematic added may help you see what im doing, I'm using the Ne5532 opamps, sorry i read now thats what you asked for. http://www.ti.com/lit/ds/symlink/ne5532.pdf
    Im usnsure how to simulate the single stage dc offset in LTspice, i was hoping that there was a way but i've searched for awhile and haven't found a method i could interpert.
    Yes cancelling the DC offset would be better than the caps if at all possible!

    Thanks again guys, this is all really helpful information.
     
  11. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Also Audioguru, its for reproduction on a full custom built sound system if it ever was to come to it so i would want to be going for at least a strong 25Hz if possible.
     
  12. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    I keep messing this up aha, Im basing it off the Ne5532s but they are not in the simulation its just a standard one that im using? I guess i should use a similar one but what will that change, if i still cant meaasure it.
    Also the ESP schem:
     
  13. Audioguru

    Expert

    Dec 20, 2007
    10,454
    1,151
    The NE5532 is old and was invented before computers and simulations were available.
    I was going to ask you why the input resistor value is so low at 22k before you said you are using these old opamps that have a minimum input resistance of only 30k ohms. I would use a modern OPA134 single, OPA2134 dual or OPA4134 quad that has extremely high input resistance Jfets on the inputs.
     
    Sulkyoptimism likes this.
  14. Veracohr

    Well-Known Member

    Jan 3, 2011
    649
    98
    Using a NE5532 or similar opamp, the value of R14 (on the non-inverting input of U1A) should match the overall resistance on the inverting input. The input bias currents to the opamp inputs develop voltages across the resistance, which becomes offset at the output. If the resistance at the two inputs is the same, the offset is mostly balanced out.

    I simulated the full circuit with NE5532 opamp models and saw only 44mV offset, which I don't find to be a big deal. But if you drop R14 to 22k (to match R2), the output offset goes down to 4mV. Dropping it a little further reduces the offset even more, because R2 isn't the only resistance affecting the inverting input.

    However, notice that the opamps in the ESP schematic are TL072. These have essentially no input bias current, so the offset isn't affected by the input resistance. Same with the models Audioguru provided, only the TL072 is quite a bit cheaper.

    If you want to look at the offset, just look at the average level at the output. Or you can just disconnect the input source and see the DC output level.
     
    Sulkyoptimism likes this.
  15. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Ah right okay, so i should up it to the Op134s? I'll do that, i guess i'll just have to find another project for the bag of Ne5532's i've got aha!
    I notice the input resistance on the datasheet for the 134s is like 10 trillion or something crazy? So i'd be okay upping the resistance.
    On that not is there anything you would suggest when looking at the ESP diagram? Am i increasing the input resistances to like a Mohm or? Because again would that not affect the signal? I'll sim it tomorrow and have a look.

    Ahh i see so its the input bias current? I note the value on the OP314s is 100pA is that a good rating? I'll have a fiddle with it and see if it changes anything on this version and the one with the new opamps.
    I've tried disconnecting the input source but that doesnt seem to give me any value, -infdB everytime. Sorry im not the best at LTspice yet, the average level of the output is currently in Amplitude over frequency, is there a different function i need to run, i've tried transient response too but it didnt yield much.

    Thanks again guys.
     
  16. Audioguru

    Expert

    Dec 20, 2007
    10,454
    1,151
    If you increase the resistance then you can decrease the coupling capacitor value by the same amount. But do it only to reduce the size and cost of a film capacitor, not megohms, do what my sketches show.

    The TL07x audio opamps are also fairly old. I used them for all my audio projects 35 years ago. They have a serious problem: Phase Inversion. If an input voltage gets within a few volts from the negative supply then the output suddenly goes as positive as it can causing severe distortion. The OPA134 and dual and quad are much better (lower distortion and noise) and do not have this problem.
     
  17. Sulkyoptimism

    Thread Starter New Member

    Sunday
    10
    0
    Right but your suggestion is to do away with the NE5532s? The OPA314's are a lot more expensive and i cant find a simulation of them for LTspice, I like the idea of better quality but i need to remain within budget as well as be able to predict the results accuratly.

    For reference how would i measure the DC offset in real life? Could i just measure the DC voltage over the capacitor at different frequencies?

    Also sorry what sketches? Is there a link i can follow?
     
  18. MrChips

    Moderator

    Oct 2, 2009
    16,817
    5,160
    Attempting to measure the input offset at the input is going to alter the offset with your meter. Measure the DC voltage at the output of the opamp instead.

    OPA134 has OFFSET TRIM pins. Connect a 100kΩ trimpot between pins-1 and 8. Connect the pot wiper to V+. (See OPA134 data sheet).
     
  19. Audioguru

    Expert

    Dec 20, 2007
    10,454
    1,151
    I did not say OPA314, instead I said OPA134. Yes, they are expensive like a cup of coffee at a restaurant.
    The NE5532 is cheap but its noise level is not spec'd. The more expensive NE5532A has guaranteed low noise.

    I thought you wanted modern high fidelity, not cheap old NE5532 opamps then spending a lot of money on huge expensive 10uF film capacitors.
    Due to input offset voltages, you do not know which polarity to connect cheap electrolytic capacitors.
    Due to a minimum input resistance of only 30k you need low value resistors and high value capacitors.

    You can build a circuit with cheap old NE5532 opamps and cheap 10uF electrolytic capacitors then measure the polarity voltage on the capacitors then correct it if needed.
    You can also measure the input resistance of hundreds of cheap old NE5532 opamps and pick ones with a input resistance much higher than 30k then use better lower value film capacitors that polarity does not matter.

    If you change the circuit to work with a single positive supply then the 10uF electrolytic coupling capacitors will always have the correct polarity and cheap old NE5532 opamps will be fine. The output might make a POP sound when turned on.

    The TL072 dual opamp is high fidelity and since it is used in millions of stereos its cost is much less than a single NE5532. The Phase Inversion problem never occurs when playing recordings that never have the signal level too high (unless you turn it up too high). Its input resistance is extremely high so you can use higher resistor values and lower value film capacitors.

    The DC offset voltage is DC only, frequency does not change it.
     
  20. Audioguru

    Expert

    Dec 20, 2007
    10,454
    1,151
    OOPs, the NE5532 is a dual opamp.
    Here is my sketch that did not become attached before:
     
Loading...