I suspected it might help a lot. The cap as it charges pulls current "out" of the opamp input pin. That appears as a huge imbalance in the two bias currents while the cap current flows. I suspect its to analyse because the exact currents will vary as the rails rise too. I would consider and try a FET opamp.
I've just been thinking of the input bias current as a constant current regardless, I guess that's kind of a simplistic approach 
Worst case input bias current for the NE5532 is 800nA, with a resistor of 10kOhm, this gives 8mV at the non-inverting input. Assuming that this voltage is the same for both inputs (so that the offset voltage is close to zero), the voltage seen on the non-inverting input at start-up must be 0V (because it starts charging the input cap) which should lead to something like -100mV at the output at start-up. I assume that we are not talking about just 800nA here? Is there anything I can do to counter this in some way? I can reduce the capacitance of the input cap, perhaps I can add a input resistor to increase the charge time of the cap or something also?
I'm going to order some fet op-amps, seems like a really good idea
Worst case input bias current for the NE5532 is 800nA, with a resistor of 10kOhm, this gives 8mV at the non-inverting input. Assuming that this voltage is the same for both inputs (so that the offset voltage is close to zero), the voltage seen on the non-inverting input at start-up must be 0V (because it starts charging the input cap) which should lead to something like -100mV at the output at start-up. I assume that we are not talking about just 800nA here? Is there anything I can do to counter this in some way? I can reduce the capacitance of the input cap, perhaps I can add a input resistor to increase the charge time of the cap or something also?I'm going to order some fet op-amps, seems like a really good idea
The offset is pretty low in practice Andrew, I've certainly never felt a need to trim it lower, even using cheap TL071 type devices. As you say, its the input bias current that is removed, and that makes a huge difference in circuits like this.
Alright, seems like lowering the capacitance of the input cap and adding a 1k input resistor just about removed the pop sound. I'm now down to about 700mV pop on one channel and less than 200mV on the other channel. It's audible on my 32Ohm cheap high sensetivity headset, but not really a big deal. It's barely audible on my AKG K-701s.
About offset voltage and input bias current. I know that a fet input op-amp will still have a offset voltage. I see that the OPA2604 has a offset voltage of 1mV TYP which isn't such a big deal, but what if I have a fet op-amp with a offset of more than 5mV? With a gain of 13 it leads to an output offset of 65mV. With the NE5532 I can have the input resistor as a pot and then 'use' the input bias current to counter the input offset voltage by turning the pot. I've done this in the past to get a offset voltage of close to 0V. With a fet input op-amp, there is nothing I can do about the offset voltage other than adding a huge cap on the output Isn't this kind of a disadvantage when dealing with fet input op-amps?
Edit: Also, my offset voltage is now down to about 500mV on both channels, not sure what I did to make that happen
About offset voltage and input bias current. I know that a fet input op-amp will still have a offset voltage. I see that the OPA2604 has a offset voltage of 1mV TYP which isn't such a big deal, but what if I have a fet op-amp with a offset of more than 5mV? With a gain of 13 it leads to an output offset of 65mV. With the NE5532 I can have the input resistor as a pot and then 'use' the input bias current to counter the input offset voltage by turning the pot. I've done this in the past to get a offset voltage of close to 0V. With a fet input op-amp, there is nothing I can do about the offset voltage other than adding a huge cap on the output
Isn't this kind of a disadvantage when dealing with fet input op-amps?Edit: Also, my offset voltage is now down to about 500mV on both channels, not sure what I did to make that happen
Last edited:
You can minimise the offset on the 5532 by equalising the bias currents between the two inputs. Is that what you mean about the pot ? That's easy to do by selecting suitable resistors for the feedback and input networks.
Fit a FET opamp and the problem will go away. 500mv is pretty high for a DC offset, even more so considering its a headphone amp. Try a TL072, cheap as chips
Fit a FET opamp and the problem will go away. 500mv is pretty high for a DC offset, even more so considering its a headphone amp. Try a TL072, cheap as chips
Another thing. OPA2604 is almost 6 times more expensive than NE5532, what am I actually paying for here? I know fet inputs is fine and dandy, but as I pointed out above, I will probably be left with a even high offset voltage on the output. I'm sure the OPA2604 has lower distortion, but it's not like it's impossible for the NE5532 to be made to have inaudible distortion either.
You are paying for a less popular device. The 5532 is used in everything, market saturation, prices are low. The 2604 is a very fine device though and in the scheme of things, all build costs considered, its a small premium to pay.
Can you not use a lower supply and then you could test a TL072 etc and see ?
Can you not use a lower supply and then you could test a TL072 etc and see ?
I guess I could lower the supply voltage just for a test, but the next step down would be to use a +-12V transformer instead of a +-15V and this would be too low for the final product.
Looking at some other op-amps I found the TLE2142. Input bias current of TYP 8nA (25 times less than the NE5532) and a TYP offset voltage of 220uV (about half of the NE5532). This seems like a good candidate
Looking at some other op-amps I found the TLE2142. Input bias current of TYP 8nA (25 times less than the NE5532) and a TYP offset voltage of 220uV (about half of the NE5532). This seems like a good candidate
I wouldn't run it at +/- 20V either but it should work ok at that voltage.
+/- 20V is a suboptimal voltage, better to go down to +/-18V since the selection of suitable OpAmps is a lot bigger at that voltage level.
Dear Fast Eddie,
I am in the process of making a "single supply version of TGM 8" . Can you kindly detail how to build a "soft start" circuit in?
http://www.diyaudio.com/forums/solid-state/250380-biguns-tgm8-single-supply.html
Thanking you,
--gannaji
To be absolutely certain that it doesn't clip when driving 600Ohm headphones. This amp can currently output around +-16.5V which gives 226mW into 600Ohms. There are probably headphones that will need more than that at peak moments when playing music, I'm not sure.
I'm still curious about my input capacitor though. I used a 4.7uF cap at first and then reduced it to 2.2uF (which drastically reduced the pop sound). I know 4.7uF is a lot for a input cap, but they don't cost more than 2.2uF so I just thought why not
When I first finished the amp I saw that the offset voltage was higher on one of the channels, I increased the input resistor from 9.1k to 13k and this made the offset equal on both channels. Since the pop sound is louder on one of the channels despite the input caps being the same, I think it has to do with the input resistors being different in value (9.1k vs. 13k). If this is the case it would mean that the pop sound relies heavily on the values of the input capacitors, input resistor to ground and perhaps a resistor connected between the input cap and the non-inverting input. I would really like to understand whats going on here, is it possible to predict how high the output peak during start-up is going to be? Here's a tiny schematic of the components in question:An externally hosted image should be here but it was not working when we last tested it.
I just wondered on the rails
The value of input cap really depends on your wanted LF response and wanted input impedance. 9.1k and 2.2uf is around 8Hz, so probably low enough.
You have to many variables to easily predict the switch on thump imo. The bias current of the 5532 would need to be measured because it determines the current that flows out of the opamp input pin. As you know looking at the data sheets, it varies from device to device by a wide margin. Also the setting of the volume control will alter the charge rate of the cap.
The value of input cap really depends on your wanted LF response and wanted input impedance. 9.1k and 2.2uf is around 8Hz, so probably low enough.
You have to many variables to easily predict the switch on thump imo. The bias current of the 5532 would need to be measured because it determines the current that flows out of the opamp input pin. As you know looking at the data sheets, it varies from device to device by a wide margin. Also the setting of the volume control will alter the charge rate of the cap.
What is the maximum voltage specified by the headphone manufacturer?
I have seen 5Vpk on some higher impedance phones.
It already has one, consisting of R4-R5-C7. Increasing the value of C7 will make it bias up slower or "softer."
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.