Fixing Phono Preamp Distortion due to Impedance Nonlinearity
I'm building a phono preamp, and trying to find a way to reduce the distortion (and noise) due to the impedance non-linearity at the inputs. This discussion is primarily about MM cartridges which have a high impedance that rises to 10's of thousands of Ohms at high frequencies due to the inductance. Also for this discussion, ignore any RIAA feedback networks and assume a flat gain stage.
Using the usual voltage amplifying input stage with negative feedback, you want the feedback resistance very low to reduce noise, so the - input typically sees less than 100 Ohms resistance. This impedance mismatch will cause a form of common mode distortion, typically close to 0.1% at high frequencies with almost all opamps. For examples showing this, see Samuel Groner's "Operational Amplifier Distortion" document, specifically the "Input Impedance Linearity" graphs, available on his web site:
SG-Acoustics · Samuel Groner · IC OpAmps
Or, see Figure 6-40 on page 6.53 from Analog Devices "Op Amp Applications" document, available from their web site.
ADI - Analog Dialogue | Op Amp Applications Handbook
(sorry I'm not able to copy images from these documents)
Here are some possible solutions:
1. Match input impedances. Insert an impedance network in the feedback leg to match the input impedance (MM cartridge plus load). See figure below. Obviously not a very practical solution that will match all cartridges, but will a network with "typical" values for a MM cartridge be better than nothing? Believe it or not, some preamp products have incorporated actual cartridges into the feedback path to reduce distortion! Again, not very practical in my opinion.
2. Bootstrapping. For example, see Analog Devices AN-232. Very interesting approach, but I am leery of a method that bootstraps the IC substrate, that is, it appears to modulate the minus supply with a version of the input signal. It is claimed to be highly effective, but I wonder what kinds of side effects it may have? Anyone try this? And what if I want to use external transistors at the inputs to reduce the input noise, will bootstrapping work for that plus the opamp too?
3. Instrumentation Amp configuration with balanced inputs. See example of classic 3 Opamp instrumentation implementation below. This approach is supposed to reduce common mode noise, but will it also cancel out non-linear common mode distortion? I have not seen any distortion plots for instrumentation amps, so I don't know how effective this approach will be. Anyone know?
Are there any other approaches? (Don't tell me to stick to MC cartridges.)
a discrete front end can have a bootstrapped cascode diff pair which largely eliminates the effect
the dielectric isolated cascode input of the OPA627 is a jfet monolithic op amp example from the opamp_distortion paper with low input impedance distortion
a simple circuit condition is maximize the supply V to reduce the input Z modulation
the nonlinear cm input impedance distortion mechanism at mm phono cart signal levels is not likely to be easily measurable:
the opamp_distortion paper uses 100 K input Z and unity gain in the nolinear input impedance test circuit, I can't find the test level explicitly stated for the input impedance test but +20 dBu seems to be the nominal test source setting
for mm phono cart the signal level is ~100x less and the source Z is at least 10x less than the test Z, for predominantly 2nd order distortion the expected distortion is directly proportional to level so the best estimate is that this distortion would be 1000x less than the plots
with the -60 dB correction for application's <100 mV input swing and <10K input impedance imbalance the AD823 (as a typical fet input op amp) input distortion would be below the noise floor in the opamp_distortion paper plot - and the OPA627 input impedance distortion is over an order of magnitude lower under the test conditions
a low noise op amp like the AD743/745 will have larger input device C and proportionately larger input nonlinearity but even at 10x greater than the AD823 or OPA134 input impedance distortion it would hugely less than any of a number of phonograph mastering, cutting, playback distortion mechanisms
output loading distortion can easily be larger with the low Z feedback required for low noise - so a output buffer in the loop is necessary to get to where its worth worrying about input nonlinear C
added loop gain in the multiloop output buffer like Jung shows improves the transfer linearity too
Have you looked at the designs of Douglas Self for an ultra low noise front end ?
IMO Dougs design is about as good as gets and addresses many issues and problems.
Around page 53 here,
Self on Audio - Google Books
Thanks guys, for your inputs.
What I want to do is use external JFETs to reduce both the input voltage and current noise. The problem is these low noise JFETs have very large capacitances, (yes, cascoding will help, but not eliminate the capacitance) more than ten times than those found in typical op amps, so I fear that the distortion effects will be magnified compared to the op amp distortion tests. Unfortunately most JFETs don't have access to the substrate for bootstrapping, though the obsolete 2SK389 and I think the LSK389 do have substrate pins.
I don't agree that a MM phono input's impedance is less than 10 times the 100K Ohms used in the Opamp test setups. Most MM's have very high inductance for example AT155LC about 500mH, or Stanton 681 910mH, resulting in around 30K Ohms at 10-20KHz (even with a 47K load in parallel), not that far from the test setup (those high impedances are a reason for my wanting to use JFETs for lower current noise). I do agree that the input levels will be smaller for a phono cartridge, on the order of 100mv peak, so that will help minimize the distortion. Yes, yes, phono distortions are gross, on the order of 1%, but that's no reason to use a poor opamp design for a phono stage, I would like to reduce the distortion as much as possible if only for my own peace of mind.
I have found another example of boostrapping to reduce distortion, by Dimitri Danyuk published in Electronic Design -
Supply Bootstrapping Reduces Distortion In Op-Amp Circuits
This is also discussed in another old thread.
I have received several responses questioning whether MM cartridges really do have impedances on the order of 30K Ohms at high frequencies. You can look up the equations for calculating R,L,C impedances in any introductory electronics or physics textbook. Plug in the numbers for a MM cartridge, typically R(DC) from 500 to 1000 Ohms, L from 500 too 1000 mH, and C(load) typically 200 to 300 pF recommended, plus 47K R(load), and you should get in the neighborhood of 30-40K Ohms at 10KHz (in answer to a question, the capacitive load helps to dampen and flatten the frequency response peak, but does not cancel out the impedance at 10-20KHz; if it were large enough to do so, the cartridge's high frequency response would be seriously reduced). See for example Dennis Colin's preamp article from audioXpress, an extract quoted below -
"I measured the impedance of a Shure R27E cartridge. Although old and inexpensive,its impedance is typical: 635Ω + 631mH. Paralleled with this preamp’s 47k5 load, the impedance seen by the preamp at 1kHz is 943Ω (resistive) + j3k84 (inductive); impedance magnitude is 3k95. But at 10kHz, z = 19k5 + j23k; magnitude is 30k2.
I had first tried an AD797 op amp alone (no JFET) at the input. With a low source impedance such as an MC cartridge provides, all was fine. But with the MM cartridge, the noise was about 8dB higher than with the JFET input; plus (rather, a negative) the noise peaked around 10kHz. Too much and hissy! The reason is the bipolar op amp’s input current noise. Now, 2pA/√Hz might not sound like much, but when multiplied by the loaded cartridge’s 30k impedance magnitude at 10kHz, a noise voltage term of 60nV/√Hz undesirably appears! This is 10.5dB higher (at 10kHz) than the 18nV/√Hz noise of the loaded cartridge’s 19k5 resistive component. Combined, the resulting NF is 10.8dB at 10kHz. The JFET reduced the NF to 0.3dB."
On another topic, here is Mike P's view on the drawbacks of bootstrapping the supply rail, taken from the other DIY (Simple Machines) Forum's thread on Sam Groner's
OpAmp Measurement Series
"there is a lurking problem with almost every version of modulated rail or bootstraped cascode distortion reducing mechanism I have studied, ugly overload characteristics. Even if care is taken to keep the circuit from going into oscillation or outright blowing transistors, you still seem to always end up with at least spikes of current on the supply rails as protection diodes start to conduct. and there is the sudden shift in conditions caused when a rail driver clips and the voltage across the active device starts to change with the signal. imagine a capacitance multiplier for each rail, with the shunt capacitors connected to the output of a common mode voltage follower instead of ground. When the followers clip It doesnt cause gross distortion but it far from ideal. this is not easy to indicate with a clip LED and changes with gain."
the opa627 still looks like the input impedance nonlinearity is going to be below the noise floor with 30 KOhm source peak using the opamp_distortion paper's numbers
bootstrapping `~100 mV signal in a gain ~10 circuit with +/15 V supplies is not going to have the issues of clipping a bootstrapped unity gain follower against the supply rails so I think as long as startup issues are controlled there is little chance of signal related bootstrap clipping issues
Dimitri's Fig 3 ps bootstrap op amp U2 will never clip, if U1 clipping causes excess ps pin current that draws down C9,10 unacceptably then adding shunt regulation (Zener or TL431) with enough bias current could eliminate the consequences)
Thanks for pointing out the opa627, which looks like it uses bootstrapping to control the impedance linearity distortion. I don't know if it will have distortion below the noise floor - below 1KHz, yes, but above, it's hard to tell from the graph (see green trace below) since it doesn't seem to specify the signal level. At any rate, the distortion seems fairly low compared to other op-amps. The big drawback for me is that the opa627 costs about US$25 each in small quantities (ouch!). At that price I think I'll try to roll my own bootstrapping circuit. I just wish that there was more information on bootstrapping - all the published circuits I could find don't give a lot of DIY design information (how much bootstrapping to use for different gains or supply voltages, what if you use discrete transistors, what value current sources or voltage regulators to use for the Danyuk circuit, etc). It seems those articles were written for expert analogue designers!
any composite op amp circuit is pretty rare despite Jerald Graeme and Walt Jung's including chapters on multiloop amps in their books
op amp supply bootstrapping is doubly rare - scattered articles, a few per decade mean the technique is not widely known or used
in diy audio there is an additional bias towards using discrete in any application - and certainly when there is any question about op amp performance limitations in the application
circuits similar to the audioXpress article's discrete fet input working inside the feedback loop with an op amp are probably the most used class of discrete/op amp hybrid composite amplifiers
|All times are GMT. The time now is 11:49 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio