I always include reverse polarity rail clamps. The other rail can easily get hit with reverse polarity transient during startup and shutdown of the power supply by any rail-to-rail parts. Any resistive voltage divider between rails or ICs between rails. Many voltage regulators don't like opposite polarity on their outputs, even for an instant, and the damage can be cumulative over time. Two diodes are cheap insurance to failure proof the design.
Last edited:
Exactly. I have even used Schottky diodes for this - they have a lower forward voltage drop and thus protect the ICs better.I always include reverse polarity rail clamps. The other rail can easily get hit with reverse polarity transient during startup and shutdown of the power supply by any rail-to-rail parts. Any resistive voltage divider between rails or ICs between rails. Many voltage regulators don't like opposite polarity on their outputs, even for an instant, and the damage can be cumulative over time. Two diodes are cheap insurance to failure proof the design.
I must agree, reverse polarity condition can happen.
May be I'm too brave or just lucky, but what I remember from my initial research about the TPS7A30/TPS7A49 regs is that they have all kinds of protections built-in. I also trust TI's evaluation board design which does not have a single protection diode and I doubt that they forgot them. I personally use four of these regs to power my op-amp based buffers for about half a year now and they run without a problem without protection diodes. On my L78xx based reg designs I have two protection diodes always included as datasheet suggests.
Regards,
Oleg
May be I'm too brave or just lucky, but what I remember from my initial research about the TPS7A30/TPS7A49 regs is that they have all kinds of protections built-in. I also trust TI's evaluation board design which does not have a single protection diode and I doubt that they forgot them. I personally use four of these regs to power my op-amp based buffers for about half a year now and they run without a problem without protection diodes. On my L78xx based reg designs I have two protection diodes always included as datasheet suggests.
Regards,
Oleg
The datasheet for these excellent regulators still show that absolute maximum ratings for the output as -0.3V and +0.3V for the +ve and -ve regulators respectively. That is an IC substrate thing - very little can be done about that. These are new devices, so I don't expect that a momentary excursion into forbidden territory will cause the device to latch up like some earlier devices would, but it is still good practice to guard against it.
The real problem with two independent regulators for the two rails is that one regulator doesn't know what the other is doing, i.e. they are not tracking regulators. Expecting the rise and fall times to be identical is somewhat over-optimistic.
Furthermore, even if the regulators themselves survive, the components comprising the load might not. E.g. for the OPA1622, the -ve rail is not supposed to go more positive than ground by 0.5V.
The real problem with two independent regulators for the two rails is that one regulator doesn't know what the other is doing, i.e. they are not tracking regulators. Expecting the rise and fall times to be identical is somewhat over-optimistic.
Furthermore, even if the regulators themselves survive, the components comprising the load might not. E.g. for the OPA1622, the -ve rail is not supposed to go more positive than ground by 0.5V.
Hi SamAnytime,
If I understand you right you are talking about some sort of fault condition. In normal operation the regs should start slowly and nearly simultaneously due to soft start effect of the noise reduction capacitors. I am still not convinced that the problem exists but for the peace of mind I'll try to fit protection diodes from rails to GND. It would probably be enough to remove a bit of solder mask to make the SMD pads for the diodes. I'll see where it can be easily done.
Note that on all implementations (I've seen at least 4) of the TPS7A47 regs there are no protection diodes and so far nobody complained about regulator failure. And the regs I use are from the same family...
Regards,
Oleg
If I understand you right you are talking about some sort of fault condition. In normal operation the regs should start slowly and nearly simultaneously due to soft start effect of the noise reduction capacitors. I am still not convinced that the problem exists but for the peace of mind I'll try to fit protection diodes from rails to GND. It would probably be enough to remove a bit of solder mask to make the SMD pads for the diodes. I'll see where it can be easily done.
Note that on all implementations (I've seen at least 4) of the TPS7A47 regs there are no protection diodes and so far nobody complained about regulator failure. And the regs I use are from the same family...
Regards,
Oleg
The load seen by each power rail is different. Even if everything is precisely symmetrical, the load seen by each rail depends on the state of the output of the opamps that are driving the headphones.
Let us suppose the outputs are being diven negative when the power is turned off. That means the -ve rail will decay before the +ve rail, creating a condition that could cause the -ve rail to be pulled +ve, above ground.
In a low power headphone amp scenario, there isn't much likelihood that serious damage will be done. Putting in those diodes, however, is good engineering practice; it won't do any harm in this case, and it might save your bacon when you're dealing with beefier circuits with massive capacitors across each power rail.
You mentioned fault conditions. Lets suppose the -ve supply dies, leaving the -ve rail floating. This will result in the -ve rail being pulled +ve which is not good for the opamp.
The reverse scenario is equally applicable. LEDs to indicate power status are also a good idea, so the ciucuit doesn't cook in a faulty condition.
Let us suppose the outputs are being diven negative when the power is turned off. That means the -ve rail will decay before the +ve rail, creating a condition that could cause the -ve rail to be pulled +ve, above ground.
In a low power headphone amp scenario, there isn't much likelihood that serious damage will be done. Putting in those diodes, however, is good engineering practice; it won't do any harm in this case, and it might save your bacon when you're dealing with beefier circuits with massive capacitors across each power rail.
You mentioned fault conditions. Lets suppose the -ve supply dies, leaving the -ve rail floating. This will result in the -ve rail being pulled +ve which is not good for the opamp.
The reverse scenario is equally applicable. LEDs to indicate power status are also a good idea, so the ciucuit doesn't cook in a faulty condition.
It will be difficult to fit SMD pads for the protection diodes on the PCB. I guess the best way is to add the diodes across the regulator's output capacitors. As for the LEDs - it's fun to see the light when something is working but human is too slow to react in the case of fault so I do not see the help here except aesthetics
Regards,
Oleg
Regards,
Oleg
I just looked through the TI's support forum regarding the protection diodes for the TPS7A30/TPS7A49 regs. Indeed, some people report failures without the reverse current protection diodes due to driving the Vout pin significantly above the Vin potential. In the case of my OPA1622 HeadAmp design this should not happen due to little capacitance after the regs compared to the filter caps on the reg’s input.., unless the input of the reg is deliberately shorted to GND. At the same time there were reports of positive regs (TPS7A49) not starting properly in bipolar supply designs because the negative regs (TPS7A30) starting first and pulling the Vout pin of the positive reg below GND potential. On small transients (depends on the soft start time constant) this would probably be partially remedied by the built-in ESD protection diodes which would of course slowly die with time since they have no capacity to withstand such transients regularly. So a pair of Schottky diodes across the rails with low Vdrop would help here as suggested above. There were also hints from the TI’s engineers to use the reverse current protection diodes in series with Vin just before the input capacitor. This increases the total Vdrop but prevents the degradation of PSRR and regulation if parallel diodes (from Vout pin to Vin pin) are used instead.
I hope I did not miss the meaning of what I read at the TI’s forum
Thanks SamAnytime for rising this question!
Regards,
Oleg
I hope I did not miss the meaning of what I read at the TI’s forum
Thanks SamAnytime for rising this question!
Regards,
Oleg
Received the boards for the buffer between the volume pot and the OPA1622 headamp. Now I have to decide if I should use LME49720 or OPA1612 as the buffer... The LME49720 I have at hands and they are not as pricy as the OPA1612. With the DC blocking capacitor between the pot wiper and the buffer I should not have any benefit from lower bias currents of the OPA1612 or there is something else to consider when deciding between these two op-amps?
I haven't used either of those as a buffer before. Why don't you try both and see if there's any difference in sound?
Obviously, lower input bias current op-amps are easier to use in DC coupled designs. With that being said, you may be able to use certain unity gain stable op-amps(especially ones with JFET inputs) without any coupling caps on the pots' wipers.
I don't like coupling caps. In fact, I usually don't build most discrete designs that use them, especially designs that have them in series with the outputs. I would much rather build a similar design that uses a servo to control the offset.
Instead of being limited to the '49720 and '1612, why don't you try the OPA1642 without the coupling caps?
It's the best sounding JFET op-amp I've tried.
Obviously, lower input bias current op-amps are easier to use in DC coupled designs. With that being said, you may be able to use certain unity gain stable op-amps(especially ones with JFET inputs) without any coupling caps on the pots' wipers.
I don't like coupling caps. In fact, I usually don't build most discrete designs that use them, especially designs that have them in series with the outputs. I would much rather build a similar design that uses a servo to control the offset.
Instead of being limited to the '49720 and '1612, why don't you try the OPA1642 without the coupling caps?
It's the best sounding JFET op-amp I've tried.
Hi ammel68,
Thanks for suggestions! I am currently assembling a list of op-amps I would like to try. I'll place and order at mouser around this weekend. I will probably assemble a set of buffers and try them. Honestly I do not expect to hear a difference unless I screw up the op-amp compensation and/or supply bypassing. I'll report about my progress/impressions here.
Regards,
Oleg
Thanks for suggestions! I am currently assembling a list of op-amps I would like to try. I'll place and order at mouser around this weekend. I will probably assemble a set of buffers and try them. Honestly I do not expect to hear a difference unless I screw up the op-amp compensation and/or supply bypassing. I'll report about my progress/impressions here.
Regards,
Oleg
There are bias current compensation schemes you might want to try. Basically, you inject a current of equal and opposite magnitude into the node in question, cancelling out the bias current. This works best if the bias current is constant over temperature and time. A 20 turn trimpot and a couple of resistors are all that is usually needed.
I just checked the datasheets of OPA1612 and LME49720 again and LME49720 has actually lower bias/offset currents. I must have looked in the wrong datasheet before and got a wrong impression. So I guess I'll skip OPA1612 for now. Also OPA1642 has rizing THD+N at higher frequencies with increasing source impedance which will interact with the volume potentiometer. So I guess I'll try LME49720 as a buffer before trying anything else. I am not afraid of using DC blocking capacitors, so I should be fine with it.
I've used such a scheme with bipolar op-amps to reduce the offset. It works quite well, but some folks are troubled about injecting any PS nasties directly into one of the op-amp's inputs.
Unfortunately, I've found some of the best sounding op-amps out there can have quite high input bias currents.
You're quite welcome.
You may want to shoot johnc(John Caldwell) at TI a PM asking what op-amps he recommends for what you're doing before placing your Mouser order.
LME49720 is a fine op amp and I don't think you'll be disappointed with it.
With regards to the OPA1642 distortion with higher source impedances, it's actually one of the best performing op amps I've ever measured in that regard and will outperform the LME49720 with high source impedances.
The old datasheet for the OPA1642 didn't show how good the performance was, and an improvement was made to the process after the datasheet was released. The new datasheet reflects how low distortion will stay even with a 10kOhm source impedance
See figure 37: http://www.ti.com/lit/ds/symlink/opa1642.pdf
Either op amp is a great option, enjoy your project!
With regards to the OPA1642 distortion with higher source impedances, it's actually one of the best performing op amps I've ever measured in that regard and will outperform the LME49720 with high source impedances.
The old datasheet for the OPA1642 didn't show how good the performance was, and an improvement was made to the process after the datasheet was released. The new datasheet reflects how low distortion will stay even with a 10kOhm source impedance
See figure 37: http://www.ti.com/lit/ds/symlink/opa1642.pdf
Either op amp is a great option, enjoy your project!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.