Thanks. Still interesting, although I'll probably stick with massively paralleled JFETs for low source Z.
It occurred to me the other night that if the source impedance is so low as to shrug off almost arbitrarily high capacitive loading, the falloff in gm with drain current being what it is and the noise voltage going inversely as the square root, one could parallel devices to get down to quite low levels. Of course you'd get there more quickly spending more current, but the idea that phantom power could be used to power hundreds of paralleled BF862s for example... It would be an odd-looking assembly in any case.
I misunderstood Dave's RF jargon (my days as an RF guy ended in the mid-70s, never looked back) and thought of the "other" Norton. As Emily Litella used to say, "That's very different. Never mind."
Groner has been getting some interesting results with paralleled 862s recently. There's a boy to watch
. We were both a little puzzled about some of the claims that there might be some sort of "thermodynamic" limit to such reduction with paralleling. He moved more quickly than I (which is not saying much ) to disprove this empirically. IIRC he's using 220nH in each gate lead.Of course it goes without saying (so why did I say it?), it's easy to mess up such front end performance in the second stage, not to mention with the power supply and about any dang thing else
Hi,
There is SO MUCH suboptimal with this from a high performance view, I would not know where to start.
Op-Amp's and full passive EQ generally do not mix well (despite the repeated publications of such circuitry, it seems the "designers" all failed some basic electronics classes or simply copied of each other without ever properly analysing these "designs"). There is a 34dB+ overload margin penalty with this design over active EQ.
If we are going to use feedback and op-amp's, why not at least make sure the feedback factor stays largely invariant across the audio bandwidth and HF distortion is lowered by NFB? Why insist on perpetuating utter sub mediocre performance designs?
With an LT1115 you can make a fine single stage "simple" MC Phono (60dB Gain) that outperforms a lot of very expensive Phono's out there both objectively and subjectively. For MM an OPA637/OPA827 is an excellent choice with similar results. In either case < 0.01% THD at any sensible output level below +20dBu, very high overload margins at any frequency if using at least +/-15V rails.
If we must absolutely have passive EQ because of some fear of negative feedback (which is silly, as the above suggesed passive EQ Phono has ton's of it anyway), then why not use some of BB's "transconductance cell" diamond transistors in parallel into a pure shunt passive EQ, at least we will not loose our overload margin as badly even if they can only take +/-5V (this only gives around 10dB overload margin penalty over active EQ on 15V rails).
There are so many ways to innovate a least a little bit and avoid poor performance, there really is no excuse for such designs in this day and age.
Ciao T
As for your "yet another phono stage", may I suggest this:
This appeared in 1990 in a German magazine. The author is one Holger Hermann, who works for Burr-Brown Germany, hence the BB op amps.
There is SO MUCH suboptimal with this from a high performance view, I would not know where to start.
Op-Amp's and full passive EQ generally do not mix well (despite the repeated publications of such circuitry, it seems the "designers" all failed some basic electronics classes or simply copied of each other without ever properly analysing these "designs"). There is a 34dB+ overload margin penalty with this design over active EQ.
If we are going to use feedback and op-amp's, why not at least make sure the feedback factor stays largely invariant across the audio bandwidth and HF distortion is lowered by NFB? Why insist on perpetuating utter sub mediocre performance designs?
With an LT1115 you can make a fine single stage "simple" MC Phono (60dB Gain) that outperforms a lot of very expensive Phono's out there both objectively and subjectively. For MM an OPA637/OPA827 is an excellent choice with similar results. In either case < 0.01% THD at any sensible output level below +20dBu, very high overload margins at any frequency if using at least +/-15V rails.
If we must absolutely have passive EQ because of some fear of negative feedback (which is silly, as the above suggesed passive EQ Phono has ton's of it anyway), then why not use some of BB's "transconductance cell" diamond transistors in parallel into a pure shunt passive EQ, at least we will not loose our overload margin as badly even if they can only take +/-5V (this only gives around 10dB overload margin penalty over active EQ on 15V rails).
There are so many ways to innovate a least a little bit and avoid poor performance, there really is no excuse for such designs in this day and age.
Ciao T
I'm not saying it's a revelation, John, I am simply saying that one has had a long and fruitful life and has pleased many an audiofile. Above all, it's simple to build, and it does have a passive eq.
As I stated before, it will sound better if discrete transistors are added as current boosters - but then, virtually all op amps will sound better that way, no matter what their specs on available current say.
It can additionally be modified as a two stage affair by splitting the rails, allowing the OPA 37 to stay at +/- 15V regulated by a pair of zeners, while for the second stage, an OPA 2604 can be used, again with discrete transistor current boosters, but at +/- 24V. This will significantly increase its overload margin AND give it a better yet sound - I find that OP 2604 has a very warm and natural sound, as opposed to later BB's offerings (after it was acquired by TI and became "The Tucson Operation"), with FET inputs, which I find to be very shrill and irritating.
Ultimately, I am not advocating this particular approach, it has simply proved its worth over the years, I think.
Thorsten, I never claimed this to be anywhere near optimal - but I do say that I have seen and heard at least 15 versions of that schematic, all of which did a good job and in fact did sound better than most other run-off-the-mill fare.
We seem to be drifting towards the exalted here. Which is fine, but there are a lot of folks out there who barely know how to use a soldering iron. They need something relatively easy and simple to build to improve on their usually factory made units, many of which are more complex yet providing worse results in audition.
It's nice when you have a device with a dual FET input duff pair, followed by a dual transistor cascode, followed by a currrent mirror, etc, etc, but one should always remember that such parts are not easy to come by and are not cheap to acquire. Individuals are not companies, which have access to stocks individuals can only dream of.
On a purely personal note, I would rather use discrete components than op amps. Op amps are a blessing to me working as protection circuits, LED array displays and DC servos, but that's about it.
We seem to be drifting towards the exalted here. Which is fine, but there are a lot of folks out there who barely know how to use a soldering iron. They need something relatively easy and simple to build to improve on their usually factory made units, many of which are more complex yet providing worse results in audition.
It's nice when you have a device with a dual FET input duff pair, followed by a dual transistor cascode, followed by a currrent mirror, etc, etc, but one should always remember that such parts are not easy to come by and are not cheap to acquire. Individuals are not companies, which have access to stocks individuals can only dream of.
On a purely personal note, I would rather use discrete components than op amps. Op amps are a blessing to me working as protection circuits, LED array displays and DC servos, but that's about it.
Hi,
I have probably heard more than 15 Variations.
They get better if the second stage gets the LF EQ wrapped into the feedback loop, instead of passive EQ.
However I find mixed EQ somehow seems to lack a certain hard to describe quality (coherence?), or it is simply that we still vary feedback factors in the audio range with passive/active mixed EQ and FUBAR the HF overload margin (unless we use John Curl's Vendetta Frontend or some similar open loop transconductance circuitry).
Once we go to a straight active EQ (4 * 10nF Polystyrene plus 100K & 8K2 plus 100R/10R to ground) AND an OPA with high AC performance (LT1115 for MC with 10R and OPA637 with 100 for MM) all things click into place and we have an even simpler design, with superior objective and subjective performance.
Extra buffers may or may not be needed on the output, it depends on a wide range of factors...
For a buffer after an Op-Amp the simplest solution is an IRF710 and an LM317 as CCS on 3W worth of heatsinking each, with 15V rails that allows us around 200mA Iq and full output into 75 Ohm.
Bigger heatsinking may be needed if we want to drive headphones, or we may back off the Iq to 40mA, leave heatsinks off entierly and have only the ability to drive around 400 Ohm to full output.
As my gear is always build "tube compatible" (>10K input impedance) I usually do not need extra line drivers, so I tend to leave them off.
Hardly, just very basic, common sense and sound (pun intended) design using Op-Amp's. Exalted would look different, very different.
Such a device is for example called OPA637 and is easy to acquire, not cheap I grant, but worth the money, used for a 40dB Gain MM Phono.
For MC the AD797 (if you have more luck keeping it stable than I had) or LT1028/1115 make excellent choices for a "single stage, single op-amp" MC Phono.
The parts needed for my "AAP21K" (read - Analog Addicits Phono for the 21st Century) can be readily obtained from either Farnell, Mouser and similar outlets, at moderate, in not dirt cheap prices. Non of it is seriously "exotic". Non of it is seriously expensive.
The key difference between the "AAP21K" and the original "AAP" is that it drops the MC Pre-Pre (with it's oh so controversial PSU filter choke) and insteads comes in 40dB Gain MM Version (OPA637 or similar Fet Op-Amp) and 60dB Gain MC version (LT1115/LT1028).
It also re-adjust the RIAA EQ Network to use four equal size (10nF), as decent quality EQ cap's are now very hard to come by and often carry larger MOQ's for individuals at places like farnell etc., so one stereo Phono can use most of the usual 10 packs of LCR made polystyrene Cap's (last pair is used to bypass the output coupling cap).
On a personal note, I would rather use Tubes than op-amp's, discrete or monolithic, but that is another kettle of fish. As would be a discrete Op-Amp version of the "AAP".
Ciao T
I have probably heard more than 15 Variations.
They get better if the second stage gets the LF EQ wrapped into the feedback loop, instead of passive EQ.
However I find mixed EQ somehow seems to lack a certain hard to describe quality (coherence?), or it is simply that we still vary feedback factors in the audio range with passive/active mixed EQ and FUBAR the HF overload margin (unless we use John Curl's Vendetta Frontend or some similar open loop transconductance circuitry).
Once we go to a straight active EQ (4 * 10nF Polystyrene plus 100K & 8K2 plus 100R/10R to ground) AND an OPA with high AC performance (LT1115 for MC with 10R and OPA637 with 100 for MM) all things click into place and we have an even simpler design, with superior objective and subjective performance.
Extra buffers may or may not be needed on the output, it depends on a wide range of factors...
For a buffer after an Op-Amp the simplest solution is an IRF710 and an LM317 as CCS on 3W worth of heatsinking each, with 15V rails that allows us around 200mA Iq and full output into 75 Ohm.
Bigger heatsinking may be needed if we want to drive headphones, or we may back off the Iq to 40mA, leave heatsinks off entierly and have only the ability to drive around 400 Ohm to full output.
As my gear is always build "tube compatible" (>10K input impedance) I usually do not need extra line drivers, so I tend to leave them off.
Hardly, just very basic, common sense and sound (pun intended) design using Op-Amp's. Exalted would look different, very different.
Such a device is for example called OPA637 and is easy to acquire, not cheap I grant, but worth the money, used for a 40dB Gain MM Phono.
For MC the AD797 (if you have more luck keeping it stable than I had) or LT1028/1115 make excellent choices for a "single stage, single op-amp" MC Phono.
The parts needed for my "AAP21K" (read - Analog Addicits Phono for the 21st Century) can be readily obtained from either Farnell, Mouser and similar outlets, at moderate, in not dirt cheap prices. Non of it is seriously "exotic". Non of it is seriously expensive.
The key difference between the "AAP21K" and the original "AAP" is that it drops the MC Pre-Pre (with it's oh so controversial PSU filter choke) and insteads comes in 40dB Gain MM Version (OPA637 or similar Fet Op-Amp) and 60dB Gain MC version (LT1115/LT1028).
It also re-adjust the RIAA EQ Network to use four equal size (10nF), as decent quality EQ cap's are now very hard to come by and often carry larger MOQ's for individuals at places like farnell etc., so one stereo Phono can use most of the usual 10 packs of LCR made polystyrene Cap's (last pair is used to bypass the output coupling cap).
On a personal note, I would rather use Tubes than op-amp's, discrete or monolithic, but that is another kettle of fish. As would be a discrete Op-Amp version of the "AAP".
Ciao T
Sorry, I forgot that there is another quite different circuit also called a Norton amp. Clever people should only be allowed to invent one circuit each!SY said:
Just so more people don't get thrown off the scent: by "Norton" I do not mean a current based op amp but a common gate/base low noise RF preamp with inductive feedback. It happened to be in my mind because of a recent discussion I had on a radio forum. Transformer quality could be an issue. In RF a wideband transformer may only have to cover one or two decades in frequency. For audio we need three decades as a bare minimum.
The Norton has some similarities to the output stage with transformer feedback to the cathode, but the Norton takes this to the ultimate in that the feedback is so heavy that it defines the stage behaviour rather than merely modifying it.
Thorsten, in general, how do you feel about an LT1115 active eq circuit, but with a gain set to a lower value, followed by another ciruit, op amp or discrete, which could run off greater voltage rails, and which would bring the output of the first stage to its full nominal value?
Have the LT do the eq part and deliver say 25 mV of linear signal, subsequently amplified times 4?
Wouldn't this help increase the LT's overload point by quite a bit, thus avoiding the problem of relatively low power rails of +/-15V?
Have the LT do the eq part and deliver say 25 mV of linear signal, subsequently amplified times 4?
Wouldn't this help increase the LT's overload point by quite a bit, thus avoiding the problem of relatively low power rails of +/-15V?
Hi,
This would in practical terms not be constructive.
To lower gain would mean to either reduce the RIAA Network Impedance which would in turn increase the loading on the output, or to increase the foot resistor in the divider, which in turn would increase the noise.
There is an optimum solution there somewhere and for standard values of the EQ Capacitors it happens the one I suggest.
At best reducing the gain of the circuit will reduce THD at the cost of noise, which seems a poor trade-off.
Anyway, the LT1115 for example has (relatively) poor common mode and output linearity, so it likes to work into relatively high impedance loads (relative to 600 Ohm) and with low input levels. Hence it is ideal for our application, even if the circuit appears to simple to be good.
We do not have any overload margin problems with the 15V rails and active EQ, they only happen with passive EQ.
For active EQ, 60dB Gain and +/-15V rails and 0.5mV @ 5cm/S we get 500mV Peak at "0dB" so we have around 28dB Overload Margin for all frequencies on the output of the phono stage that produce an equal 500mV peak.
So if gain is set correctly there is no overload margin issue. Incidentally the same applies to the MM version.
For passive EQ we must either compromise overload margin or noise over optimum solutions.
Ciao T
This would in practical terms not be constructive.
To lower gain would mean to either reduce the RIAA Network Impedance which would in turn increase the loading on the output, or to increase the foot resistor in the divider, which in turn would increase the noise.
There is an optimum solution there somewhere and for standard values of the EQ Capacitors it happens the one I suggest.
At best reducing the gain of the circuit will reduce THD at the cost of noise, which seems a poor trade-off.
Anyway, the LT1115 for example has (relatively) poor common mode and output linearity, so it likes to work into relatively high impedance loads (relative to 600 Ohm) and with low input levels. Hence it is ideal for our application, even if the circuit appears to simple to be good.
We do not have any overload margin problems with the 15V rails and active EQ, they only happen with passive EQ.
For active EQ, 60dB Gain and +/-15V rails and 0.5mV @ 5cm/S we get 500mV Peak at "0dB" so we have around 28dB Overload Margin for all frequencies on the output of the phono stage that produce an equal 500mV peak.
So if gain is set correctly there is no overload margin issue. Incidentally the same applies to the MM version.
For passive EQ we must either compromise overload margin or noise over optimum solutions.
Ciao T
The overload margin is really not that bad, though it won't have the stunningly good numbers of a tube circuit.
The concern is really the first gain stage, since following the EQ, the high frequencies are attenuated, and the output of a cartridge at low frequencies is... low. You've got a gain of 45, so if we assume that the opamp can swing 13V peak with the 15V supply, that's 9.3VRMS. Divide by the stage gain and the input overloads at something like 200mV. That's 32dB above rated cartridge output for most MMs (5mV). Now certainly cartridges output more than their rating when hitting ticks and pops, but 20-25mV is about the most you're likely to see. So the overload margin is not state-of-the-art, but probably good enough.
On topic, my own subjective and unsupported experience is that phono stages with high overload margins make poor records more listenable. How high is high? That's a different question, but I think that 200mV is good enough.
I seem to remember a Pioneer integrated amp, a long, long time ago, in another galaxy ... which had +/- 53V rails, or some such very high number. I remember thinking that was enough to do 100W/8 Ohms.
I also seem to remember reading somewhere, really a long a time ago, that "crackle" and "pop" on LP records could get as much as 200 mV to the input stage. Under normal conditions, for MM cartridges, that would be 20...30 mV.
I also seem to remember reading somewhere, really a long a time ago, that "crackle" and "pop" on LP records could get as much as 200 mV to the input stage. Under normal conditions, for MM cartridges, that would be 20...30 mV.
There's a boy to watch
Sam and I are good buddies
That comment might have come from me and it got slightly misconstrued. I think it was MTS instruments (they make giant counterweights to put on top of skyscrapers for active damping) that found for low frequency noise in air there seemed to be a point where the heat flow from the increased power started to increase the 0.1 - 10Hz noise to counteract device paralleling.They used strain gages to sense building flex and I guess the natural freqency of a 100 story tower is well below a Hz.
Sam and I are good buddies
They used strain gages to sense building flex and I guess the natural freqency of a 100 story tower is well below a Hz.
If it had its origin with you, that wasn't attributed. Where I saw it was in the pages of Jan's bookzine, in connection with others' phono preamp pieces. About the limits to noise performance, it was a speculation with arm-waving mostly, and in one case provoked a LTE from me (about that and many other issues in the particular article, which was something of a tutorial) and one from Burkhard Vogel, who, as is his wont, filled several pages with equations
I still was unable to persuade the original author that using a resistor for feedback around an inverting amplifier did not lower its current noise, and I was advised to do a simulation The skyscraper thing reminds me of the story about Tesla, which I suspect is a very old urban legend, about his determining some natural resonance of a building and installing a little stimulator in it, which according to legend was about to cause the structure to collapse before he turned it off.
If I wanted to bring a structure down, I'd just arrange to rent out my upstairs neighbors in the apartment here
At least they don't play music --- perhaps they are stomping around with iPods or the ilk.Those are two different clever Nortons. The one after whom the current-differencing amps are named (but certainly not their inventor) was Edward Lawry Norton, 1898-1983.
I believe this effect is called sympathetic vibrations. If memory serves, the idea is that if you can get the natural resonance of a building to coincide with external vibrations, they would multiply and the amplitude would be quite destructive.
If I wanted to bring a structure down, I'd just arrange to rent out my upstairs neighbors in the apartment here
Brad, you want to get them before they multiply.
- Status
- Not open for further replies.