John,
Thanks,
I admit there can be some confusion from my picking the second harmonic as an indicator. The mechanism that creates the second harmonic distortion most likely in my OPINION is a very low voltage dead zone.
This will also create IM (intermodulation distortion) that is much more annoying than just the second harmonic. I actually would not expect clean second harmonic to be annoying below a few percent.
The other issue I forgot to mention was that the energy in music is not uniform with frequency, so going from 250hz to 500hz would probably add another 4db or so to the perception of distortion.
So if someone does not understand that the distortion that is just measurable is related to something harder to measure they might chase the wrong source.
ES
No, with a 300 ohm source MC that would be 6db or the distortion numbers would be 1/2 what is given. But with a typical 30 ohms that would only be a 10% change.
Since these numbers are in my OPINION +/- a factor of 10, that size error is on the list of a few other factors I did not include for clarity, as most of them tend to cancel.
But it certainly is a factor.
Halcro have been using SMPS's for years although you wouldn't call them
mainstream.
T
Please demonstrate -34dB distortion from a "bad" cartridge loading resistor. You just gave a senario for its existance. There are test records that can do better than 2%. Better yet just use an old cart as a source impedance with a signal generator. I also think you need to recalculate the effect of resistor non-linearity working against the cart impedance.
BTW a dead zone makes odds and at some low input there would be no output or 100% distortion. In 35yr. of reading The Journal of the Instrumentation and Measurement Society I have never seen mention of any of these effects, and these folks are looking at pico-volts.
A fixed level independent component of distortion is probably non-physical. I would look more to measurement artifact. For seconds flipping the resistor should flip the phase. Any fixed component would imply distortion with no input, this case is obviously thermodynamically impossible.
By not exhaustively exploring cause and effect in mesurements one shows experimenter bias.
. Why not to try to find the real issue?
Here is another radical approach that works for power factor correction (specifically harmonic reduction):United States Patent: 5691577.
I have used it with exceptional results for the audio quality. However it makes for heavy power supplies, since you need heavy iron to get it to work. Few transformer guys can even make these inductors. Years ago I negotiated a license but could never get a cost effective product going.
Ahhh! Didn't know that. No mainstream, I agree, but proof that it CAN be done well.
jd
Scott,
I am happy to play here. I am sharing stuff as I encounter it, that is why I thank you for your rigor.
I will try using a 30 ohm coil loaded by one of the bad carbon comps at 300 ohms into an equalized preamp (as soon as I get the chance to build one). The input to the coil will be magnetic to limit my variables. Output voltage from the coil should be around 100nv (F=250hz) to simulate the level from an MC cartridge. I will aim to see if the actual measurement is -46 as there will be no weighting to account for Fletcher Munson effects. As I understand the use of the termination resistor is plugged into a discrete RCA plug jack combo so I will include that. The test resistor will be free air.
I am treating the cartridge reactance as if it is in series with the voltage source.
As to the source of distortion I thought I made it clear this is at the limits of measurement and is based on a model to explain the results.
I am not so sure reversing the resistor will change the phase as I suspect at these levels I may be seeing thermocouple effects rather than barrier punch through.
I am curious as to how you measure picovolts as when I look at connectors there seem to be some that have problems at tens of millivolts.
Comments are always welcome.
ES
No problem, you may not guess but I would love to find one of these effects. As an instrumentation problem, you could use a lock-in amplifier and huge FFT's rather than a phono stage.
As for these effects please consider things like strain guages, not particularly good resistors operating with 10V excitation. People like Metler use them in scales with ppm resolution at 30mV full scale.
It is routine to measure potentials of single cell activities. CCD cameras shuffle signals through 1000's of stages (with really "bad" poly resistors in places) without losing an "electron". A cooled CCD astrometric camera can have < 1e rms noise. The list goes on.
RIAA eq only uses 6db/oct slopes. Distortion at 500hz will be cut by about 3dB (not boosted by 12dB) relative to the fundamental at 250hz.
Distortion due to non-linearity of the 300 ohm load resistor will be reduced ten-fold (not 10%) when driven from a 30 ohm source.
This 20dB difference plus the 15dB difference from the RIAA calculation above suggest that the distortion you're looking for would be about 35dB lower than you expected.
Surely you mean 100uV, not 100nV?
No problem, first things first. I have a source for WWII NOS carbon comp resistors, I hope for them to be sufficiently bad. You need to get out of the audio mindset, in other fields these measurement are routine.
Actually, I think it's the other way around. If you have (say) a 300R loading resistor that arguendo has some second order distortion and a 30R source cartridge (more likely these days to be 3R), the variation of resistance of the parallel combination of the load resistor and cartridge is reduced by about 10x, no? And with a more modern 3R cartridge, reduced by a factor of 100?
Yes, seems I got the EQ backwards only 1 pole between the 2 and a pole not a zero.
The output of the cartridge can be treated as voltage source with a resistor in series or a current source with the resistance in parallel. So we disagree on that.
Yes 100uV.
Thanks
Last edited:
Scott,
Actually I have built strain gauge systems. My experience is that with the ones I used even though designed for excitation at 10 volts anything above 2.5 volts was enough to cause self heating and throw off the measurements.
I will use an actual phono preamp as anything else would leave someone able to claim the ether was perpendicular to the astral plain or some such.
Oh by the way old resistors seem to mellow a bit and work better, new ones from the same manufacturer seem to be worse! Of course if you can get some that have been stored in a moist environment they might be worse.
ES
One of the research products on my test bench was to confirm the results shown in the current issue of AX showing input output comparisons on two different amplifiers.
I hooked up two gain clones. One to an 8 ohm 200 watt load bank (4 2 ohm 50 watt metal cased resistors on a heat sink), the other to an 18" woofer (half of an EV subwoofer system). I used a transversal filter oscillator that has 128 steps for each cycle of a sine wave. I did not use a lowpass smothing filter as I want the steps to show.
I looked to see if any of the individual steps changed length. I was able to set the trigger level to look at each step. On the resistor bank each and every step was the same.
I compared the outputs of both amplifiers in the CH1-CH2 mode and could see just at 1/10 of a division or so a sine-ish wave of the same frequency. My math function did not allow amplification so I will repeat this experiment with my AP S2.
I did not see any distortion in the waveform until the amplifiers clipped. I was unable to confirm the great changes shown in AX. I believe what was shown is the gain clone clipping with a DC offset.
I used a test waveform of 59hz. This gives a very easy to hear indication when there is a problem with power supply ripple.
Oh when I looked very carefully at the step size of the amplifier driving the subwoofer by turning up the sweep speed so I got each step almost the entire screen width and checking each step and overlaying it on a stored sample I found that from zero to 360 degrees some steps varied from others in amplitude as expected.
Oh before I forget when I first tried the test I forgot to adjust the test generator for 0 DC offset. This made the Gain Clone very rich in even order harmonics. I suspect that is why some folks enjoy it. It also came with a feedback resistor that is not among the best but is higher in even order harmonics than most other metal film types.
But what was interesting, at least to some is how long I can stretch out reporting the result is I saw no change in the timing!
I hooked up two gain clones. One to an 8 ohm 200 watt load bank (4 2 ohm 50 watt metal cased resistors on a heat sink), the other to an 18" woofer (half of an EV subwoofer system). I used a transversal filter oscillator that has 128 steps for each cycle of a sine wave. I did not use a lowpass smothing filter as I want the steps to show.
I looked to see if any of the individual steps changed length. I was able to set the trigger level to look at each step. On the resistor bank each and every step was the same.
I compared the outputs of both amplifiers in the CH1-CH2 mode and could see just at 1/10 of a division or so a sine-ish wave of the same frequency. My math function did not allow amplification so I will repeat this experiment with my AP S2.
I did not see any distortion in the waveform until the amplifiers clipped. I was unable to confirm the great changes shown in AX. I believe what was shown is the gain clone clipping with a DC offset.
I used a test waveform of 59hz. This gives a very easy to hear indication when there is a problem with power supply ripple.
Oh when I looked very carefully at the step size of the amplifier driving the subwoofer by turning up the sweep speed so I got each step almost the entire screen width and checking each step and overlaying it on a stored sample I found that from zero to 360 degrees some steps varied from others in amplitude as expected.
Oh before I forget when I first tried the test I forgot to adjust the test generator for 0 DC offset. This made the Gain Clone very rich in even order harmonics. I suspect that is why some folks enjoy it. It also came with a feedback resistor that is not among the best but is higher in even order harmonics than most other metal film types.
But what was interesting, at least to some is how long I can stretch out reporting the result is I saw no change in the timing!
Once again, let's talk about resistors in general. Resistors have been made for electronics for a very long time, perhaps 100 years. During the '20's to the '60's, most common resistors were made of carbon. They had a precision between 5% and 20%. Most of the time, they worked OK, except for EXCESS NOISE. This is the extra noise that is triggered by DC (usually) across the resistor. Some brands, like Allen Bradley or AB appeared to be superior to most 'common' carbon resistors, in regard to excess noise, and were used by serious audio professionals. For example, if you took an Ampex tape recorder apart that was made in the '60's, you would find AB resistors virtually everywhere.
While 1% metal film resistors and precision wirewound resistors were available, they were ONLY used when required, such as in precision pro video applications.
Dick Burwen, based on his experience with discrete op amp design with ADI, convinced Mark Levinson (the man) to use precision 1% resistors in his audio products. At first, they seemed ridiculously expensive, but with quantity purchases, etc, Mark found it possible. This changed audio, almost from the first. Especially when we replaced quality IC chips with class A discrete designs. We wound up 'on top of the world' in the audio community, almost instantly, and we didn't even use tubes. When you get this sort of response, internationally, you realize that you most likely have done something right. It now extends beyond measurement, but what people really hear.
At this time, the exact brand of 1% resistor was not considered very important. We used IRC, Dale, Corning, etc, interchangeably. This can be shown by an old photograph of a circuit module in an early Levinson product.
Unfortunately, resistor manufacture changes with time. Some products are discontinued and others added, and sometimes for the worse. This happened to me when Corning dropped the 'grey' resistors and replaced them with 'green' resistors.
However, thanks to Scott Wurcer helping me measure the differences with my Quantech (Scott used to be interested in such things) I was able to measure the difference in EXCESS NOISE, between the two resistor types. For the record, I was specifically told by the factory that there would be no change. So much for application engineers. But I digress.
The most important thing is that not ALL precision resistors measure the same.
Now, besides noise, what about distortion? Well, there are differences there also, even though they appear to be small differences with parts that many of our critics would not even bother to source. Now, if the expensive and established brands have differences, what about the cheap parts from Korea or China, etc? If you measure one, and it is OK, can you depend that all the others are just as good? What are the symptoms of a 'bad' resistor? Noise, distortion, both?
Well, one of the best ways, beyond serious listening, is to characterize the resistors with measurements that we CAN make with today's audio test equipment, and ferret out the exceptions to the ideal. This will help to make better audio designs, be them amateur or commercial, and avoid the GM effect of fixing the problem after the design is built, but instead, being more like Honda of QC up front.
The same philosophy can be used with resistor selection.
While 1% metal film resistors and precision wirewound resistors were available, they were ONLY used when required, such as in precision pro video applications.
Dick Burwen, based on his experience with discrete op amp design with ADI, convinced Mark Levinson (the man) to use precision 1% resistors in his audio products. At first, they seemed ridiculously expensive, but with quantity purchases, etc, Mark found it possible. This changed audio, almost from the first. Especially when we replaced quality IC chips with class A discrete designs. We wound up 'on top of the world' in the audio community, almost instantly, and we didn't even use tubes. When you get this sort of response, internationally, you realize that you most likely have done something right. It now extends beyond measurement, but what people really hear.
At this time, the exact brand of 1% resistor was not considered very important. We used IRC, Dale, Corning, etc, interchangeably. This can be shown by an old photograph of a circuit module in an early Levinson product.
Unfortunately, resistor manufacture changes with time. Some products are discontinued and others added, and sometimes for the worse. This happened to me when Corning dropped the 'grey' resistors and replaced them with 'green' resistors.
However, thanks to Scott Wurcer helping me measure the differences with my Quantech (Scott used to be interested in such things) I was able to measure the difference in EXCESS NOISE, between the two resistor types. For the record, I was specifically told by the factory that there would be no change. So much for application engineers. But I digress.
The most important thing is that not ALL precision resistors measure the same.
Now, besides noise, what about distortion? Well, there are differences there also, even though they appear to be small differences with parts that many of our critics would not even bother to source. Now, if the expensive and established brands have differences, what about the cheap parts from Korea or China, etc? If you measure one, and it is OK, can you depend that all the others are just as good? What are the symptoms of a 'bad' resistor? Noise, distortion, both?
Well, one of the best ways, beyond serious listening, is to characterize the resistors with measurements that we CAN make with today's audio test equipment, and ferret out the exceptions to the ideal. This will help to make better audio designs, be them amateur or commercial, and avoid the GM effect of fixing the problem after the design is built, but instead, being more like Honda of QC up front.
The same philosophy can be used with resistor selection.
Last edited:
When it comes to general design, I would like to point out, as I have done over the years, that there is more to conventional audio design than just traditional measurements. I did not make this up for some obscure reason. 40 years ago, I would followed what many here feel is important, namely traditional measurements, and ignored my ears when they contradicted them. Well, I got some good advice at that time, about 1970 from the late Richard Heyser. He explained what had surprised him, deviating from measured norm. He was also the poster boy for time delay minimization, when everybody thought he was over the top. He was the first to tell me about finding a highly distorting resistor in an amp design when he was at JPL. This is essentially the same mechanism the Ed Simon is measuring today, just a much worse example. This is what put me on the path, 40 years ago, along with the so so response that I was getting from rock musicians regarding my designs for them, to find the hidden factors that we were not measuring. At the time, Matti Otala came to mind, and I followed his lead. At the time, I could measure down to .005% SMPTE IM. Good enough for just about anybody, yet it was still hit and miss with making what other people felt were quality audio designs.
However, by refining my discrete designs to include jfets, and using Matti Otala's recommendations of low feedback, high open loop bandwidth, and lower order distortion by excluding class B output stages, I convinced both the Grateful Dead, and Mark Levinson that I had found the right path. Neither exactly knew what I was doing, but they both liked the results. This is what I have pursued since the early '70's, and since it works, I hope to continue doing it with gusto. I welcome any audio related input that encourages making better designs, both measurably and subjectively. Negative comments like pity for my beliefs, I can do without. OK?
However, by refining my discrete designs to include jfets, and using Matti Otala's recommendations of low feedback, high open loop bandwidth, and lower order distortion by excluding class B output stages, I convinced both the Grateful Dead, and Mark Levinson that I had found the right path. Neither exactly knew what I was doing, but they both liked the results. This is what I have pursued since the early '70's, and since it works, I hope to continue doing it with gusto. I welcome any audio related input that encourages making better designs, both measurably and subjectively. Negative comments like pity for my beliefs, I can do without. OK?
Last edited:
- Status
- Not open for further replies.