This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

OPA1622: As single supply transformer driver

Part Number: OPA1622
Other Parts Discussed in Thread: INA1620, , BUF634A, OPA210, BUF634, NE5534

I'm thinking of using the OPA1622 as a parallel buffer to drive a line output transformer with a 33Ω-100Ω primary impedance and a 10dBu step up. Power is 24V single. One little PCB would do six of these outputs, and there would be six PCBs in the system. I've combined the parallel buffer from the INA1620 application note with the midsupply reference application drawing on the BUF634A data sheet. There is one BUF634A and six OPA1622s on each PCB. I doubt the semiconductor aspect of this design needs to exceed 80kHz flat bandwidth.

Some mysteries and possible adjustments below...any thoughts welcome!

  1. I'm unsure about smoothing/bypass cap values for this sort of load.
  2. Are there any supporting components I'm missing?
  3. With that 10K bias resistor, I can't tell if either half of the OPA1622 requires a resistor (or RC pair) in its feedback loop. Transients can be unpredictable (unsure if they would be more than 10mA), and I don't fully understand what happens at powerup when the amp is used as a unity gain buffer with a low or nonexistant source resistance.
  4. I'm unsure whether the ground pin ties to midsupply or 0V. I think midsupply, drawn as such.
  5. How large of a startup transient will be produced if the enable pin is simply tied to Vcc, rather than getting 22K to Vcc and 10K to ground (or midsupply here)? The data sheet talks about it being noticeable on headphones.
  6. While the BUF634A is an easy choice to buffer the mid-supply reference, I'm unsure if it's overkill -- maybe an OPA210 or similar would suffice? How much does offset matter here? The BUF634A has a large offset IIRC.

Thanks for taking a look.

  • BTW those 10n ceramics from Vcc and Vee to midsupply should say 100n, and the AC coupling into the primary is now 1000uF || .1uF.

  • Jonah,

    Thanks for providing a schematic. I would need some time to look this over but to start off, the ground pin should be tied to the system ground or the mid supply system reference point. 

    Regarding the enable pin, I am unsure how much current consumption matters for your application however this is one thing to consider. You can estimate the current consumption on the enable pin as shown below. 

  • Here, this one should be easier to read. And I updated a few things...the amp doesn't need to work that hard, and it's now a single channel PCB that sits at each transformer.

    Seems like the enable pin in this case would draw ~20uA. Nothing compared to the 1622 driving 90Ω when terminated 1K2 || 10K, or especially 33Ω when terminated 1K2 || 600R. Haven't done the current draw calcs for the 1622 in this application, and am not exactly sure how -- any guidance appreciated. Should be less than the traditional discrete op amps used for applications like this, and made more efficient due to the load sharing.

  • Fixed the power supply, I think.

  • Jonah,

    I will review your updated schematic. 

  • Jonah,

    1. I'm unsure about smoothing/bypass cap values for this sort of load.
      1. Generally adding bulk decoupling capacitance (100uF) in parallel with 0.1uF is recommended. There will be a time constant associated with the 10R 1/2W and the 100nF cap. It does look like you’ve considered this by changing the capacitor value. If my math is correct the associated frequency is now about 159kHz. The prior RC would’ve been down at around 1.6kHz which would be problematic for passing signals above. This does come at a tradeoff of having lower decoupling capacitor values. I will reach out to some team members to see if we can identify a better filter scheme here.

    2. Are there any supporting components I'm missing?
      1. You might consider adding TVS protection diodes. Please see Ti Precision Labs link below regarding this protection:
      2. You may also consider putting zero ohm place holders on the inputs of the OPA1622 devices. U1A has some protection with the 1.5kohm. During a high frequency event it could be bypassed by the parallel cap so I would recommend adding a zero ohm place holder in series with the inputs if needed. U1B has no protection from transformer flyback current. I would also add zero ohm place holders in series with the inputs there as well.

    3. With that 10K bias resistor, I can't tell if either half of the OPA1622 requires a resistor (or RC pair) in its feedback loop. Transients can be unpredictable (unsure if they would be more than 10mA), and I don't fully understand what happens at powerup when the amp is used as a unity gain buffer with a low or nonexistent source resistance.
      1. You can match the feedback resistor to also be 10k. In unity gain the device will be fine. We recommend protecting against currents greater than 10mA. You will have to make in circuit measurements on these transients. It's difficult to say how much flyback current can occur with the transformer. The zero ohm placeholders in series with your inputs will save you a board design if you find they are needed. 

    4. I'm unsure whether the ground pin ties to midsupply or 0V. I think midsupply, drawn as such.
      1. I have reached out to the designer regarding this to provide more clarity. I will update you once I hear from him.

    5. How large of a startup transient will be produced if the enable pin is simply tied to Vcc, rather than getting 22K to Vcc and 10K to ground (or midsupply here)? The data sheet talks about it being noticeable on headphones.
      1. There is already a 500k resistor in series on this pin. Adding any more resistance won’t matter much here.
      2. Can you point me to the discussion you found on the headphones?

    6. While the BUF634A is an easy choice to buffer the mid-supply reference, I'm unsure if it's overkill -- maybe an OPA210 or similar would suffice? How much does offset matter here?
      1. Most op amps cannot drive large cap loads so the BUF634 is unique here. I would say it is a good choice for providing the reference voltage with heavy cap load.
  • Thank you, Chris, for the thoughtful response.

    The following may aid in the discussion. Here is one of the later (still pre-1980!) amps that was designed to drive this transformer. And below are the transformer specs.

    IC1 above is a very early NE5534.

    Typically there is a gain of 10 at this stage by tying -ve to 0V through a resistor and large cap in series. It is possible that I will leave pads open at the inverting input of my U1A for gain. Regardless, I am still a little confused about how to safely use the 1622 in unity gain without adding much if any series resistance. If the only way to protect my U1B is to add series resistance, are we talking about 100Ω or 1000Ω here? I can certainly leave 0Ω pads on the PCB for it, and the same for U1A after the decoupling and shunting. Just wondering if the 1622 is really all that different than a conventional amp that the series resistor has to be all that large to limit current. I suppose another approach would be to simply put 1K in both U1A and U1B feedback loops, and eschew the bypass entirely. Doesn't help with THD, but might be simpler and safer. Lastly (assuming unity only), I could run U1A and U1B *actually* in parallel by tying their non-inverting inputs together, matching their feedback loop RC, and summing through 1Ω resistors as drawn. What do you think? I think this is the biggest question.

    The midsupply reference voltage in every classic application of this type of amp is unbuffered. One reason I am adding a buffer is to be able to use 100K values for my divider rather than 4K7 or 10K, which I'm assuming is more friendly on the power supply.  But the main reason I'm buffering the reference is because of the OPA1622's super high bias current, which I am assuming would be problematic with anything greater than a 10K shunt / bias resistor. Otherwise it would be immensely preferable to use 1uF coupling and 100K shunt to midsupply. Please correct me if I am thinking about this the wrong way; it would be really great to not have to buffer the reference. For now I will switch back to the BUF634A.

    I think, based on your comments, that I will raise the value of the 100uF reservoir cap, at least to 220uF. Even the size of a modern 470uF/35V (which was the traditional value for this amp stage) is not really a concern. I am however unsure whether the two 10uF X7R that I have drawn from the midsupply reference to the general Vcc and Vee supply are A) necessary at all, or B) much too small and better to be 'lytic. (?)

    Regarding the enable pin, I was referring to this voltage divider below. Is it unnecessary if the supplies are not being ramped? I've seen a TI recommendation on this forum to simply tie enable to Vcc, which is simplest. I am just trying to understand if there are any caveats to that. I can't find the post about the headphones, it's somewhere on here though. If we're talking about just a few mV, I am definitely not going to add parts to abate that.

    Thanks again for the insights, on a weekend no less!

  • To address the current limiting and flyback current concerns without adding series resistance, would this INA1620 application be a more robust approach?

  • Jonah,

    I have confirmed that the ground pin can be tied to Vee from the designer. This will be the best option since it will be the lowest impedance path to its reference point as possible.

     

    To address the current limiting and flyback current concerns without adding series resistance, would this INA1620 application be a more robust approach?

    This approach will not be more robust due to any over voltage from inductive kickback. If you are concerned about inductive kickbacks you will have to make measurements on your circuit to determine how much is being presented to the OPA1620. Please keep in mind that the INA1620 has matched resistors but can vary from 1k to 0.84k up to 1.15k.

     Just wondering if the 1622 is really all that different than a conventional amp that the series resistor has to be all that large to limit current.

    There isn’t anything different about the OPA1622. Placing Shottky diodes from input to supply line is common as all Op Amps will have some input current limitations. This is only needed if you are concerned that your circuit will present an over voltage situation. It certainly wouldn’t hurt to have place holders on the PCB with Shottky place holders from the input to supply lines and TVS protection on the supplies to ground. The series resistance is the only way to limit current into the inputs and are only needed to protect the Shottky in event there is too much current. Sizing the resistor will have to be determined based on your measurements on an oscilloscope to determine the max over voltage seen at the inputs.

    I suppose another approach would be to simply put 1K in both U1A and U1B feedback loops, and eschew the bypass entirely. Doesn't help with THD, but might be simpler and safer. Lastly (assuming unity only), I could run U1A and U1B *actually* in parallel by tying their non-inverting inputs together, matching their feedback loop RC, and summing through 1Ω resistors as drawn. What do you think? I think this is the biggest question.

    I have redrawn the circuit below with the OPA1622 devices in parallel. This would be a better solution from an AC perspective. You are essentially removing a delay from one amp to another that is presented in the series case. Also you can make each feedback loop symmetrical with the same passives. Again the passives in the feedback aren’t necessary outside of my recommendations for input protection which can be the diodes and placeholders for series input resistance if you determine it is needed.

    One reason I am adding a buffer is to be able to use 100K values for my divider rather than 4K7 or 10K, which I'm assuming is more friendly on the power supply. 

    Keep in mind that the BUF634A has about 8mA of bias current already. I have added a cap in parallel with the 100k ohm resistor for low pass filtering. This will provide a clean quiet reference to the OPA1622 inputs.

    I am however unsure whether the two 10uF X7R that I have drawn from the midsupply reference to the general Vcc and Vee supply are A) necessary at all, or B) much too small and better to be 'lytic. (?)

    No you don't need the 10nF's to the supplies. You slow down the response with these caps. It wouldn't hurt to have place holders. They seem redundant. I have kept the 10nF at the VCC line down to VEE as shown. Based on my previous comments it we recommend 100uF bulk cap at the supply and then 0.1uF as close to the dut pin as possible for local decoupling.

    If you have clean supply lines you could eliminate the buffer and use the equivalent resistor divider as shown below to generate the 12V reference point.

    I don’t think the 10R resistor is helping you and can be removed. Again leaving placeholders for unexpected things can never hurt but that resistor in particular is creating a slowed down response to the capacitor charging for local decoupling.

     

    Regarding the enable pin, I was referring to this voltage divider below. Is it unnecessary if the supplies are not being ramped? I've seen a TI recommendation on this forum to simply tie enable to Vcc, which is simplest. I am just trying to understand if there are any caveats to that. I can't find the post about the headphones, it's somewhere on here though. If we're talking about just a few mV, I am definitely not going to add parts to abate that.

    The voltage divider scheme is a good approach however I would also add a capacitor in parallel to the 10.7k resistor. This will slow down the turn on time. Essentially what you’re trying to do here is have the enable pin turn on to the required voltage to turn the amp on once the supply line hits 24V. This will eliminate any concern for audible pops or clicks. You can bump up the resistor values as well to reduce current consumption if desired.