TI E2E Community
Audio Amplifiers Forum
some OPA627 questions
Was the architecture of the OPA627 changed sometime ago? I see that the BP version has better specs all around than the AP, costs more and is less available. Does the BP version have a different design than the AP or is it simply better quality control? Will the BP stay current?
I am using the OPA627BP extensively in a 4 stage phono preamp. Recently, in addition to the second and third stages, we tried the 627 in the first stage. This stage's impedence can vary from 30 ohms to 47K ohms depending on the cartridge. We noticed that at about 300 ohms the 627 starts to feel the burden of the lower impedence and it's output is reduced proportionally to the lower input impedences.
Now, it still sounds very fine; even at 30 ohms/reduced output, but I'm wondering if I am playing with fire in the long run. I considered the OPA1611 but the 627 just sounds so good. Given that;
1. Do you think the 627 will be OK under a low impedence condition long term...and
2.Do you recommend the 1611 or another opamp that would be more acceptable of input impedences that can range from 30ohms -47kohms?
Are you there? Was the new TINA file OK? I'm still wondering if the R values around U05, and R25 are acceptable. I ask this because I see feedback resistors as high as 20K with R ground at 1K for a gain of 21 in some circuit designs and others have low values like in mine. I'm not sure when to use higher values and when the less noise?(lower values) are acceptable. It would seem that lower vales, (as long as bandwidth and stability are not compromised) would always be better. Thanks
PS If you want to talk hi-end audio on your time my e-mail is email@example.com
Sorry for the delay in my reply, I'm actually posting from a tiny little hotel outside of Frankfurt as I'm in Germany at the moment! Surprisingly I have wireless internet!
Anyway, low value feedback resistors are actually a trade-off between noise performance (which would suggest lower is better) and the loading placed on the opamp. Remember that the feedback resistors are a load on the output in addition to the load that is intended. Low value resistors for the output stage to deliver more current which can cause distortion or limited the output swing in extreme cases. A good practice would be to determine the maximum signal level expected at the output of the amplifier, and to make sure that the output is not exceeding its current delivering specifications into the parallel combination of the load and the feedback network. Or as a simple design rule, most of TI's audio opamps are capable of driving a 600 Ohm load with incredibly low distortion, just keep the parallel combination of the load and the feedback resistors greater than 600 ohms and you should be in the clear. Remember that resistors produce Gaussian or white noise, not the hum your experiencing. Below is a graph that I borrowed from one of my colleagues presentations (Art Kay):
Notice that at approximately 100 Ohms of resistance, the noise spectral density of the resistor is pretty close to the input voltage noise of the OPA1611. Reducing these values further will most likely not provide much benefit. On the other end of the spectrum, extremely high values of feedback resistors can often cause stability issues as they are large enough to interact with parasitic board capacitances as well as the input capacitances of the amplifier itself, reducing the phase margin of the feedback loop.
Analog Applications Engineer
PA Linear Apps
I hope you're seeing some sites while in Germany.
Thanks for the info. I've seen the graphs, and the formula that states Rf // Rg < 600 ohms, but my problem, having no formal training is that I'm not sure how to interpret what it means. I see that and think OK; Rf X Rg / Rf+Rg >600 Ohms, but when I do the math it means that to be "Safe" for a gain of 7V/V that Rf=4.2K and Rg=700ohms, and these values seem far to high from a noise standpoint, especially when the input is somewhere between 0.15 - 0.5 mV, hence my really low values of Rf=500R and Rg (when both are // (250R//125R=83.3R) I thought that the current from the cartridge should be small enough that I can configure the 1611 as I did. But....
Add to that the cartridge loading for these low output MC carts from 30R to 100R "grounding" the OPA1611 input, and the thing becomes a nightmare even if I know the correct formulas to apply.
I recommend to the end user to try 500R as a start for cartridge loading and work down in 100R increments, and to use the jumpers judiciously so that first stage gain settings brings the cart output up to about 1mV, which is then very easy for the remaining 3 stages to deliver quiet, clean gain of 60-70dB, but I'm worried that these low values around the first stage, as you indicated, might be loading down the 1611 when all parameters are in worst case scenario.
I can do the math if I have the formulas, and understand where and how to implement them with respect to the circuit. Unfortunately for me, much of the information provided in white papers and app. notes presumes specific prior knowledge, which over the years I have been able to figure out, but here I feel like I'm on thin ice, and need you to get me pointed to solid ground; even though it may take a while.
To start, there is nothing "magic" about the 600 Ohms number, but it's important to see where it comes from. The limiting factor here is how much current the OPA1611 is able to deliver without the output stage beginning to introduce distortion. The output current is determined by the output voltage and the parallel combination of the load resistor and the feedback network. Let's consider the "worst case" for output current, which would be the amplifier swinging all the way to one power supply or the other:
This is where the 600 ohms number comes from, at full output voltage swing, the amplifier is delivering 25mA, which is well within the capabilities of the amplifier as shown in the distortion plots. But this is not strictly valid in your application because you will never see 15V at the output of your first stage. From your post it sounds like the maximum output voltage will be 3.5mV (.5mV x 7V/V). Thus, the output current from the OPA1611 in this case is:
It's safe to say your values are well within the safe output current range of the OPA1611. You can reduce them further if you like. but the white noise produced by the input voltage noise of the OPA1611 (1.1nV * Gain = 7.7nV/rtHz at the output) is going to be dominant here in terms of white noise (remember this is hiss NOT hum).
I don't think you're excessively loading the OPA1611 at all. I think the issue is that you're picking up 60Hz noise from the cartridge, cable, or possibly even the layout, and then it is being amplified by the pre-amp. Since this is EXTRINSIC noise (comes from the outside world, not the physical nature of the components) changing the resistor values won't make any difference.
When dealing with EMI issues (electromagnetic interference) we recommend a 3 step approach to all customers. First, find the source of the noise, reducing the amount of noise it is radiating if possible. Second identify the coupling medium and reduce its effectiveness (e.g. how is this noise being converted from electromagnetic radiation to a conducted signal at the input to my circuit?). The coupling medium is often inductive (long pcb traces, large loop areas) for low frequency noise such as 60Hz hum. This stage involves improving layout and grounding, implementing shielding, exploring different cabling topologies, Finally, reduce the circuit's susceptibility to this noise. This step involves improving filtering at the front end or power supply.
Tracking down and handling 60Hz issues really takes patience, unfortunately I can't provide a formula to eliminate hum. However, this does give credence to my opinion that the output phono cartridges should use a balanced shielded cable (such as a microphone cable) instead of an unbalanced coaxial cable. I've never understood why modern turntables still use RCA cables?!
Thanks for the math. It's so easy once explained.
For a while, I taught math to College students whose second language was English. What I found in all cases was that they were failing not because they couldn't do the work but simply because they didn't get the terms or the concepts. Although the professors weren't teaching abnormally fast, to these kids, hearing it in English was like he was talking at the speed of light. Once we sat down and I explained abcissa, coordinate, trinomial, cosine, cubed root etc., and they could understand what they were and how to apply the formulas, they all went to A grade students. It's like that with me sometimes; good instincts plus good explanations equals success.
Thanks for the EMI refresher course. Now that my mind is at ease about the circuit design, I have been concentrating on a course of experiments to isolate and eliminate the hum, which is only audible at phoho stage gains of greater than 58dB.
So , you believe that all phono cartridges should be wired for 3 wire balanced L and 3 wire balanced R channels, with a differential driver input on the phono stage?
Back in the last century in my recording studio, we wired the mic's with the shield drain (foil wrap) connected only at the mixing console (star) So with that in mind would a balanced phono cart wiring require a special cartridge? I thought that the shield was internally wired at the cartridge.
I recently shorted out the phono inputs and the hum went away. It was barely audible with the system pumped and the phono pre only 12" away from my sub's speaker, so I'm rather sure that the hum induced over the cable/arm cartridge etc. I was able to lessen it considerable by fudging the turntable and cable route, but...
Now, you have me thinking about a differential input to the phono.
On pg. 13 of the OPA627/637 white paper, there are 2 circuits using the 637 / INA105-106. Regarding this design I have a few questions.
Is this circuit appropriate for a phono input differential? If yes, then...
Why the OPA637 and not the 627?
Do you think the OPA1611 could be substituted for the OPA637 or 627? (I know your feelings on bipolar amps, but still...)
Re: Cartridge loading; If needed on both the non and inverting signals at pin3 of the 637, would they need to be scaled? ie: for a loading of 100R in an unbalanced system woulf the 2 resistors need to be 200R each at the 637, and for caps of 50pF would they each need to be 25pFon the balanced circuit?
For the output of the INA105-106 driving the next stage, that being an OPA627 would a resistor of 50-100R be needed and/or advised?
Would you suggest an entirely different way of doing the differential?
The advantage of a differential system is that extrinsic noise can be equally coupled into both signal conductors, and will then appear at the input of the amplifier as common-mode noise. If the amplifier has high common-mode rejection, the extrinsic noise is removed from the signal path.
I should state right off the bat that a phono output is not a "true" differential system when using RCA/coaxial cables because the impedance of the inner conductor and the shield will most likely be different (measuring this on some RCA cables is on my to-do list). An impedance mismatch causes the common-mode rejection ratio of the input amplifier to be degraded but the severity of this effect may not be a "deal breaker". The INA826 has some interesting plots in the datasheet on the effects of impedance mismatch on CMRR. I would love if turntables had shielded balanced cables for the outputs but unfortunately this isn't the case.
The instrumentation amplifier design shown in the OPA627/637 datasheet is a good "textbook" implementation and may be appropriate for a phono differential input however I would personally use the INA134 or INA137 instead of the INA105 as they are designed specifically for audio use. The OPA637 is shown on the front-end of the circuit in the datasheet because it is being used in a gain greater than 5, meaning that the circuit benefits from the OPA637s AC performance without requiring the unity gain stability of the OPA627. Notice that the datasheet is advertising the bandwidth of that circuit.
One option I have been exploring is the use of a low-noise instrumentation amplifier on the front-end of the pre-amp, such as the INA163, INA103, or INA217. All of these parts offer extremely low noise and in the case of the INA103, the higher supply rails offer improved overload margin. Again, because of their bipolar inputs I should state that I feel they should be AC coupled to the phono cartridge in order to prevent their input bias currents from passing through the coil.
In terms of cartridge loading, consider the the two sets of termination resistors and capacitors to be in series, so the resistances will need to be halved and the capacitances will need to be doubled.
just had a short eye on this thread. Regarding the input bias issue of bipolar OPamps.
When I designed my bachelor thesis (a modular pramp) in the mid80s, I also wondered if a very lownoise bipolar input OPamp, hence large bias currents, could be used without probs or if a coupling cap were required.
Well, in all my phono-stages I never used a coupling cap at the inputs. I used an INA instead.While this isn´t a direct solution to the issue, I never had an issue in praxis. And man, when a 5.000 bucks costing Koetsu Urushi, or a Lyra Titan i is running in Your stage You wouldn´t want to experience such an issue ;-)
At that time there were hardly any INAs available that featured lowest noise specs and a sufficiently high bandwidth. There were the PMI/AD 2015-2017 and the then new INA103.
The stage I designed and which I still build and use with only minor changes (nearly identical stages are the Clearaudio balanced series) uses the INA103 or the newer smaller brother INA163. Its a highly flexible design, allowing to accomodate nearly every pickup, from high output MMs to low output MCs.
The input impedance may be switched between several resistive and capacitve loads. Also one can switch between balanced and unbalanced input. The gain of the complete stage may be varied from ~35dB to ~70dB.
A DC-servo (JFET-input) feeds into the Ref-pin of the INA. The servo either with a very low cutoff frequency (RIAA) or a cutoff of 20Hz (RIAA-IEC). The latter renders a additional subsonic-filter obsolete.
Following is a 2120Hz-LP, done passive with an RC-network. Via jumper one can also choose between standard RIAA and RIAA/Neumann equalization curve.
The signal then runs into a second JFET-OP (noninverting). The feedback network realizes the 50-500Hz equalization and adds some 10-20dB of gain.
If balanced outputs are wished for a driver like the DRV134 may be added.
If noise and overload requirements can be fulfilled, a design like this, a linear gain-stage, followed by a passive LP, followed by an equalizing gain stage is probabely the best and most flexible way to cope with phono signals (rem: the subsonic may either be an active dc-servo or a simple cap). The splitting of the equalization curve allows to optimize the parts values for noise and prescision.
Only if I needed to design for an MM pickup I´d choose an JFET-OP like the OPA627 as input, but surely not for an MC pickup. I´d probabely use a bipolar OPamp like the AD797 or an parallel array of discrete JFETs like LSK389 or BF862.
It sounds like a great design. Where do you build these and for whom? How strongly do you feel that the high bi-polar currents can harm MC cartridges? Are there any papers verifying or disproving this? I have been running a Sussuro into an OPA1611 for a few months; sounds great, but I am a bit worried that it will be a costly experiment.
Just to jump back in on this topic:
I don't think high bias currents will be a major concern for MC cartridges, I was more worried about their effects on MM cartridges. I should also point out that the INA103 and INA163 that Chris mentions have bias currents much larger than the OPA1611 and if he has never had an issue than you are most likely in the clear.
I'm not aware of any papers investigating this effect to date. I was planning to look into it soon (once I'm back in the U.S.) to see if small DC bias currents caused a shift in the lower resonant point of the system which would indicate a change in the mechanical compliance due to the magnetic force created by the DC bias current.
Hi,attached is the principal schematic. The circuit has been discussed in a german audio forum (AAA-forum, analog audio association) and was built under the name PlatINA. A number of my suggestions were taken over in the design and layout. So far nobody made negative experiences once he got it running. There are even some guys who said that this might probabely be their ´last´ phono circuit they´ll ever need or own.A similar discussion went on in the DIY-Audio forum. As mentioned before the quite well bespoken Clearaudio Balanced series (especially the Balanced Reference) are slightly simpler, but close copies of this design.My own designs can be found at http://www.ami-hifi.de/ and my personal site at http://calvins-audio-page.jimdo.com/
A short description:This Phono-stage is very flexible with regard to accommodation of pickups and required gain. Pickups generating voltages rom ~150µV to ~10mV (@1kHz, 5cm velocity) can be used. I don´t know a single quality pickup that doesn´t fall into this range.principle schematics:Rin+, Rin- and Cin define the input impedance and as such the cartridge loading. Rinvar and Cinvar are additional parts that may be switched on with DIP-switches to present the cartridge a optimal loading. Depending on the setting of X1 and X2 the input runs either balanced or unbalanced, depending on the used pickup cartridge and cabling. Phono-cartridges are inherently balanced sources, but most MMs are forcefully unbalanced in that one of the pins is connected to gnd. Such pickups may be tweaked balanced again by cutting the wire or metal strip to ground. MC-pickups are generally balanced.The pickup signal runs into the inputs of an INA. The INA103 or its smaller brother the INA163 are ideal candidates as they offer very low voltage noise with an optimum around 200-400Ohms of source impedance (IIRC). Wide gain range of ~40dB up to high gain factors, ranging from 10-1000, with an optimum around +40dB. This are optimized parameters for high output MCs. But they are still well enough for low-output MCs up to high-output MMs. The full power Bandwidth remains large enough even at +60dB of gain. The INA103 offers more pins than the INA163, but those are normally not used. One exception may be if You want to implement a shield driver for the input cabling. But this is rather an non issue if You already run balanced, well shielded cabling anyway and it requires another OPamp and special connectors and cables. The INA163 is absolutely sufficient for this task and considerably cheaper.The resistor RG defines the minimum gain. The parallelable resistors RGb increase the gain of the INA. Gain range copuld be between +20 and +55dB (@1kHz).Since this is a linear gain stage which is fed with a preequalized signal of raising amplitude with raising frequency, one needs to keep an eye on the possibility of overload. The preequalizing RIAA curve gives +20dB at 20kHz compared to 1kHz and further raising above 20kHz. The Opamp must cope with a 20dB-plus-a-certain-safety-margin-of-say-10dB-level, compared to 1kHz. In difference to a measuring signal music seldomly shows a high HF content, which relieves the danger of overload. Still though, You should keep it in mind. The great advantage of the linear stage as first stage is the possible high flexibility. You hardly find another Phono-Preamp with similar features if any.The JFET-Input OP OP1 forms a dc-servo or subsonic filter, depending on the cutoff frequency defined by Rdc/Cdc. The original RIAA1963 EQ-curve defines just 3 time constants 75µs (2120Hz), 318µs (500Hz) and 3180µs (50Hz), no subsonic. In that case a ´standard 1M/1µ-filter is sufficient. A latter refinement added a 4th time constant to the RIAA/IEC curvature of 7950µs (20Hz, e.g. 750K/10n). Since the Opamp sees large differing source impedances at its inputs a JFET input OP is the part of choice. Bipolars would ad offset and noise. If one dislikes dc-servos a series coupling cap at the INAs output and its Ref-pin tied to gnd is a passive alternative.This linear gain input stage is followed by a passive lowpass (RF1+RF1b||RFn / CF1,X7 closed) which creates the 75µs/2120Hz time constant. The cutterhead manufacturer Neumann introduced another timeconstant of 3.18µs (50kHz). After RIAA the pre-EQ must raise till infinite with rising frequency. The mechanical and electrical limits of the cutterheads forced a low time constant in praxis anyway. Neumann used the 50kHz network in their mashines which became kind of standard. RF1 / Rf1b+CF1 and X6 closed, X7 open, realize the so called Neumann-timeconstant. At replay the amplitude response remains linear to app.100kHz while the standard RIAA already drops above 20kHz. This results in a slight sonic difference. You can choose which one You prefer. Since the filter works by voltage divider action one should see that the ´loss´ of the filter remains low, preferrably below -1dB@1kHz.Remarks: a active non-inverting OPamp filter´s gain would eventually drop to 1, while the RIAA requires a drop down to 0. So a passive RC-network behind the OP amp would be required to correct for this ´misbehaviour´ anyway. It might also be an idea to implement the lowpass within the first gain stage, the INA, using the gain-drive pins. But that would spoil CMRR, one of the main reasons to use an INA here, since the noise figure of an INA is 3dB higher (as in any difference amp compared to a single ended input stage). The INA103 and 163 are nevertheless so good, that this Phono-Preamp is one of the lowest in noise I came across over many years of audio experience. Noise reamins under all circumstances below the -67dB physical lmit of Vinyl.A gain stage with equalizing network in the feedback loop follows the lowpass filter. Rf2, RF3, CF2-3 define the 318µs/500Hz, 3180µs/50Hz time constants. RF2-3 sets the gain. DC-gain is +20dB against the 1kHz-gain, plus the 1kHz-gain. The gain at 1kHz may be low around 3-5 times. The source impedances of the two inputs may vary and differ largely. So Offset must be kept in mind. A JFET-OPamp is the part of choice. One could add a large cap from RF2-3 to gnd to reduce the dc-gain to 1 and as such reduce output Offset. Alternatively one could add a second dc-servo. But so far I never had any issues with Offsets. HiFi is quite low-demanding in this regard. The OPs output is decoupled with a small series-R to cope easier with capacitive loads. A dedicated Buffer to drive heavier loads like the BUF634 may be used insted. If one wants balanced outputs a DRV134/DRV135 may be added.Splitting the equalization into two parts allows for independant design and easier calculation of the filters, which results in higher precision and allows to optimize the filters with regard to impedance values, losses and noise contribution. A all-in-one active eq hardly achieves better than +-1dB amplitude linearity over the full audio bandwidth. With 1% parts the split eq may even reach less than +-0.1dB. Since both OPamps should be JFET-input types one can of course use Dual-OPamps. OPA2604, OPA2134 and better parts like the OPA2827 are ok.Chrisps. the problem of dc-bias feeding into the pickup cartridge seems to be no issue in literature, cause it´s hardly mentioned anywhere. A change of mechanical compliance may indeed occur, but I assume that it´d be of insignificant amount of change. A slight increase in inertia of the tonearm would counter the effect. Adding a tiny weight of just 1 gr. on the headshell will probabely show greater effect. I was rather worried, if especially the low impedance of MC-pickups would allow for such great currents that the microscopic wire coils could be harmed (currents of several mA running over the input protection diodes? Especially when switching the phono stage On or off). But as told before, I hadn´t any pickup defect in nearly 30years.
Concerning the noise of the phono preamps, there is a tradeoff between noise voltage and noise current at the input amp. For the MC cartridges the amps with lowest noise voltage are better, the OPA211 being the best (including the distortion). For the MM cartridges the amps with low noise current and reasonable noise voltage are suited better, the OPA827 being the best choice (again including the distortion). Note I made an extensive measurements of THD for several tens of models at gain 100 and frequency 15 kHz, which corresponds with the real conditions better then the data in most datasheets. The best choice for the balanced input and MC is INA103. The really good balanced input for the MM can be built only using hybrid circuit with discrete JFET pair as front end - the best available model for this purpose is the LSK170.
BTW, if you read the diyaudio, you are common with the Actidamp concept maybe, which is my work.
I think we have all agreed on the tradeoff between voltage noise and current noise depending on the cartridge type. However, I should also point out that due to the elevated gain at low frequencies in the RIAA correction curve, the rule of thumb "FET devices are always better for MM cartridges" is not axiomatically true because of the increased 1/f noise many of these devices exhibit. Because of the design and process of the OPA1641 and OPA827 they feature far lower 1/f noise than typical FET-input type amplifiers which makes them very good choices for MM phono pre-amplifiers. Also note that a rumble filter will help with the 1/f noise issue.
Another exception to this rule is the OPA209 which features uniquely low input current noise (for a bipolar) due to the input transistor topology of the part. While experimenting in Tina with different phono circuits, I found the OPA209 provided very good noise performance with MM cartridges, however I have not listened to this part yet in person to verify other aspects of its audio performance.
The OPA211 is a good choice for MC cartridges, but I would suggest that you use the OPA1611 instead as it should have even lower THD numbers, especially at higher frequencies, while maintaining identical noise performance.
On a final note, have you compared the LSK170 to the BF862 for discrete MC headamps?
first , the OPA1611 (or 1612) exhibits virtually the same distortion figures like the OPA211, but there is pretty big spread of actual values, so the measurement of more pieces would be needed, which I did not. Frankly said, recently is a phono preamp before finish, where the OPA1611 will be used - because of lower price. The OPA209 seems to be a good compromise - I intend to test them, too.
second, the LSK170 exhibits virtually the same noise behavior as its predecessor 2SK170, including the low 1/f corner. The BF862 is very similar or better on mid and high audio frequencies. I can't say more, but I plan to do more detailed measurements in near future. It's a pity that the better types toshiba (2SK147 and equivalents) are out of production.
Hello John & Bohumil:
Of course there is a place for noise and other bench measurements, but isn't what comes out of the speakers the only real concern? In blind tests, my jury unanimously preferred the OPA627 over the 827. But that was with opamps preceding it other than the now installed OPA1611. Recently I built a version with all TCC Z-foil smd resistors instead of the usual IRC tanfilm and just this change made significant inprovement. Perhaps now reinstalling OPA 827's would make an impact better than the 627's but then going back to the other front end op amps may be better too! It's easy to see how more time can be spent on designing and building than enjoying the music.
I find it really really hard to believe that there are a few opamps that will be the best in all phono pre circuits; and if so, it's not noise and distortion measurements that are going to be the determining factors.
All content and materials on this site are provided "as is". TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with regard to these materials, including but not limited to all implied warranties and conditions of merchantability, fitness for a particular purpose, title and non-infringement of any third party intellectual property right. TI and its respective suppliers and providers of content make no representations about the suitability of these materials for any purpose and disclaim all warranties and conditions with respect to these materials. No license, either express or implied, by estoppel or otherwise, is granted by TI. Use of the information on this site may require a license from a third party, or a license from TI.
TI is a global semiconductor design and manufacturing company. Innovate with 100,000+ analog ICs andembedded processors, along with software, tools and the industry’s largest sales/support staff.