SAS Audio LabsTM
Where Music Comes Alive TM
Welcome to SAS Audio Labs. SAS Audio Labs came about in December 1996 (after 16+ years of research, 38 years now. I started in the electronics 57 years ago), and continue to play a leadership role in the evolution towards audio perfection. I hope the knowledge presented will help you in understanding the difficulties in designing a great sounding audio system, and helps you when evaluating a component.
This White Paper will cover an assortment of concepts. Each section is concise and to the point. This will include:
- The Entire Design
- Maximize Sonics, Synergy
- Feedback and Power Supply
- Impedance Mismatch and Distortion Changes
- Output Impedance Change and Frequency Response
- Hardwiring and PC Board Comparison
- Globel Negative Feedback. Very Very Carefully
- Amateur Comparison of Parts, Either Article or Forum String. Help or Hinder
- Power Transformer for Class A Operation
- Rectifier Tube VS Solid State Rectification
- Marketing Hype; Materials Do Not Matter
I start each topic with general easy to understand information. My conclusion is as non technical as possible. Almost all of the information presented is first semester, first year electronics. Some information was addressed some 50+ years ago, or earlier, in the RCA Radiotron Designers Handbook, written by 26 engineers. Yet very few, if any designers/engineers/companies address the issues now or in the past. As one reads through, notice the variety of topics and difficulties that an engineer must contend with.
For the record, I am beholden to neither recent technology nor old, do not belong to any audio organizations, nor do I belong to any buddy buddy groups, or financially attached to any particular forums.
There is alot more to designing a component than solving a few equations, running a computer program, using a certain tube, or reading a few "scientific" journals with conclusions that contain comments such as "we believe" or "it appears". The problem is that a few equations, a computer program model are not complete models. An example later, under global negative feedback.
Let us continue on, we have alot to cover.
The Entire Design is Important.
SAS Audio Labs believes that the entire design is important, which includes the power supply, the active circuitry, individual parts incorporated, and even the layout. Although a particular design or layout may look good on the surface and have excellent specifications, and a lot of support; digging deeper may reveal inherent weaknesses that renders the design or layout used inferior. For example:
Concerning layout, does it have hot parts or tubes (especially octal tubes) in close proximity to temperature sensitive parts, such as capacitors? A 250-490 degree octal tube within two inches of a 170 to 220 degree maximum rated capacitor risks failure sooner than later. For electrolytic capacitors, the life span is approximately halved with every 20 F degree rise in temperature. Blowing a power supply could cause an expensive down stream component to fail which both you and I do not want.
It has been found that a 5khz (5,000 cycles per second) or lower signal will couple to other wires or parts. Thus "crosstalk" between parts and channels is a concern. The channel separation may be 90db at 1khz but only 70db at 10khz. Now suppose we have a cymbal crash in the left channel, with harmonics to 20khz (and beyond). As a result, some of those harmonics will appear in the right channel, causing imaging problems. Secondly, the frequency response (FR) is changed due to extra output from the unwanted right channel. So although single channel measurements may indicate perfection in the lab, such is not the case in the real world.
As one can see, layout is very important indeed.
Maximizing Sonics, Synergy.
One of the mistaken "facts" in audio is that simply matching components maximizes synergy. Unfortunately, there are at least three problems with that line of reasoning.
- First, synergy is not an absolute, but variable. Remember when you upgraded a component and the system sounded better? The synergy was better, but is it the best possible? ( Now I am not saying a $10,000 preamplifier is better than a $3,000.000. However, a $500.00 preamplifier is almost always inferior.)
- Second, one can not perfectly cancel a fault. Nelson Pass stated (paraphrase), once pristine is gone, it is gone. That is very true. How does one recover lost dynamics, separation of instruments, proper tone/harmonic structure? In order to have perfect reproduction, the individual parts and components must be the best, most natural, and accurate possible.
- Third, very very rarely does a component have only one flaw, or sonic signature. Now combine components, each with multiple flaws, and it is impossible to arrive at perfect reproduction.
In otherwards, for maximum syntergy, it is better to remove as many flaws as possible from each component before creating a system.
Feedback and Power Supply.
In October, 1997, I presented a "White Paper" that discussed the issue of frequency dependent musical feedback from stage to stage, through the common power supply itself. This type of feedback is problematic in almost every single positive voltage power supply type designs, yet is almost always swept under the rug. Either the designers do not understand electronics, or they don't wish you to know. (Actually, this problem is mentioned in the "RCA Radiotron Designers Handbook," 1960, 4th edition, written by 26 engineers, so this particular issue is not new, just brought back to life.)
Larger de-coupling capacitors are sometimes mentioned as solving this problem. However, as another of my white papers explain, one is simply trading one problem for another, such as ESR and internal inductance problems become more prominent in the larger capacitor.
Musical feedback extends from near zero hertz through at least upper bass frequencies, and can occur at very high frequencies, thus its consequences are wide spread within the audio band.
Impedance "mismatch", distortion changes.
What is the effect of amplifier input impedance VS active preamplifier/source output impedance. Most recomment a 10:1 ratio. (The specifications of both impedances can be found in the owner's manuals.)
I also recommend a 10:1 ratio to be safe. (The RCA Radiotron Designers Handbook recommends a 5:1 ratio.) Using a 10:1 ratio, the amplifier input impedance should be 20,000 ohms (20k ohms) with the preamplifier output impedance of 2000 ohms (2k ohms).
However, some falsely claim/market a 100:1 ratio to "reduce distortion". For an amplifier input impedance of 20k ohms, one would need the preamplifier/source to have an output impedance of only 200 ohms. So does adding a low output impedance buffer stage actually lower distortion?
Understand that teaching the 100:1 ratio attempts to legitimize the use of a buffer stage while inferring that those who use a 10:1 ratio are inferior. Not only does adding a buffer stage not significantly reduce distortion, but deteriorates the musical quality by adding another stage with distortions of its own, increases the complexity, increases "crosstalk" problems between channels, and adds to the cost (which increases the profit margin). Let's check out an example.
We have an amplifier with 20k ohms input impedance (Z). Let's compare a preamplifier with a 100 ohm output Z to a 2000 (2K) ohm output Z to a load. As such, we are decreasing the ratio from 200:1 to 10:1.
The total harmonic distortion of a JJ E88cc tube, at 2v rms output measures approximately 0,01% (-80db) using the 20:1 ratio. Changing the ratio to 200:1 (assume perfect cathode follower) lowers the distortion by approximately 0,0012% to -81db. So the distortion lowers from -80db to -81db. A real world buffer stage, itself, would add another stage, more complexity, more distortion than the savings.
A marketing trick that sounds good, but only adds to the cost, and the manufacturer's profits.
Output Impedance Changes and Frequency Response Improvements. Does It Really Help?
This section deals with the high frequency response of our active preamplifier with and without a buffer stage. We will use a 50pf ic VS a high 250pf ic capacitance load. The output impedance with buffer stage is 100 ohms. Without is 2,000 (2k) ohms.
"udb" is micro db, or millionths of one db, "mdb" is milli db, or thousandths of one db. Z is impedance.
First, the high capacitance 250pf interconnect cable and the buffer stage, 100 ohms. The high frequency response drops approximately 44udb at 20 khz. With output Z of 2khz, the drop is 17mdb at 20khz. Not much is it.
Now we use the 50pf interconnect cable 100 ohm output impedance. The result is 1.8udb at 20 khz. With output Z of 2khz, the drop is 0,6mdb at 20khz. Again, little different. (Rarely, a longer IC with higher capacitance is neccessary and a buffer output stage is necessary)
As one can once again see, the added buffer stage also does not appreciably extend the high frequency response. Yet the additional stage adds cost while degrading the music.
So the question is, why not just design a single stage, low output impedance, wide bandwidth, and low distortion design to begin with and forget the additional buffer stage with its associated problems and cost to you?
Hard Wiring and PC Board comparison.
In December, 1996 I published a "White Paper" listing the positive and negative virtues of both Hardwiring and PC boards. The trick to using PC boards is, of course, avoiding the negatives while incorporating the positves.
Two big pluses of incorporating PC boards are unit to unit specification consistency and higher frequency response due to less stray capacitance, both between adjacent parts and from parts to ground. At the same time, exploiting the virtues of hardwiring such as very low resistance and fewer solder connections. In fact, "lead to lead" beats hardwiring at its own game.
Global Negative Feedback? Very Very Carefully.
Ok, an example from above. At the outset, unless the designer is using either very small amounts or hugely amounts of global negative feedback, higher orders of harmonics are produced. Nelson Pass has an article explaining, and using actual measurements to demonstrate this. See below link.
A second problem is the amplifier's overall time constant; delaying the signal from intput to output. A simple oscilloscope measurement with, say, a 10khz signal (open loop amplifier bandwidth 70khz), and one will discover that global feedback is delayed by many microseconds (us), not nano-seconds (ns) or "instantaneous" as some claim. At least two major studies, one with three mainstream national medical organizations and universities, demonstrate the ear is sensitive to 2us and 5us changes. By definition, any delayed and altered information is again fed back at the amplifier's input again. Computer models and mathematical calculations are not fully complete. (Watch out for the fireworks, reactions of all sorts by global feedback proponents.)
Another problem is the creation of higher order harmonics under most conditions. This is discussed and actually measured in Nelson Pass' article.
Nelson Pass, Global Negative Feedback:
The problem was mentioned in the 1952 RCA Radiotron Designers Handbook, so known for over 60 years.
Amateur Comparison Testing of Parts, Whether an Article or Forum String, Do They Help or Hinder.
With the introduction of the internet, forums were established that allowed just about anyone to pass along information from anywhere. Most of the time we know little if anything about the poster or reviewer. How a poster or reviewer's information was gathered, processed, and the resulting conclusion is sometimes rather interesting. Usually it consists simply of, "we stuck in a part and the sound was better," in a component, so no accurate or stringent listening tests were performed.
For instance, when comparing vacuum tube "T" to vacuum tube "U", is each tube operated under its optimum conditions? What design is being used? Is the design actually accurate in order to give an accurate accounting of "T" and "U"? If the design is bright, then a "dark" sounding tube will be ranked most accurate. Rankings of tubes and other parts on the internet are often suspect at best. What about capacitors?
Consider 0.47uf capacitor "A" being touted as sounding better than 0.47uf capacitor "B". One needs to ask:
Is the value 0.47uf the correct size to give an accurate representation of the music? Actually no. Almost all designs use insufficient size coupling capacitors for an accurate assessment. As a result, the darker, more inaccurate capacitor "A" is chosen, when in actuality "B" was the more accurate capacitor. Why is this important? First, one is using a part with at least one sonic signature/flaw. On page one, I explained the problems of combining components with flaws. The same problems exist when using compromised parts. The more flaws and problems, the less synergy.
Back to our capacitor comparison. If the proper value capacitor (s) were used, "B" would have been chosen over "A". Not only would the bass have been sufficient, but the midrange and highs would have been more accurate as well. However, since "A" was chosen by improper testing, "A" is on the market while "B" became extinct, and sound quality suffers.
Because of improper testing and evaluations in articles and on forums, all the accurate capacitors have become extinct. I have not and will not review the newer, very expensive capacitors since the cost of said parts is prohibitive. Needless to say, the amateur testers have just caused everyone to pay more for the same quality part. Man made inflation.
Chokes create sonic problems. Some claim that a choke is superior to a resistor. Some simply claim that chokes are solid engineering and have always been used over the years.
Unfortunately, all is not that simple. True, the DC voltage is not reduced as much with a choke. However:
How can a component sound accurate with such compromises?
- Good resistors can have virtually no "sound" of their own.
- Chokes cause eliptical loadlines, wherever they are placed in the signal circuitry.
- A choke adds its own distortions and reactance, thus artificially flavoring the sound.
- Chokes used between the "decoupling" capacitor and previous capacitor creates problems as the reactance of the choke varies with frequency. For example, at 20khz, a 5hy choke has 600,000 ohms impedance separating the two filter capacitors. At 20hz however, only 600 ohms separates the two capacitors. Besides that, there is dc resistance of the winding which makes the relationship non-linear. Though simplistic, it is as if the "decoupling" capacitor is actually changing value with frequency. Much more complex though.
It is true chokes have been used for decades, but for reasons quite different than most believe and post. Not in any particular order.
- Because they limited surge current through rectifier tubes when the circuit was powered on. There are whole sections in the RCA Radiotron Designers Handbook and other engineering books covering this very subject.
- The power transformer could be physically smaller using choke input filters, less peak current, thus saving money.
- Filter capacitors were of limited quality, so a choke was a way of reducing the number and physical size of filter capacitors, thus improve reliability.
- Since filter capacitors were also of limited physical size and uf, reducing hum was difficult. Chokes were a reliable way of helping to reduced hum at the speaker.
So while chokes performed some basic functions, improving sonic quality is another matter. Here are some other basic principles.
- A choke's magnetic field does not fluctuate instantaneously, but takes time which is dictated mainly by the core material. It is a form of electrical inertia, thus adding distortion.
Resistors, on the other hand, have virtually no such field to create distortion, and the current fluctuates nearly instantaneously.
- As mentioned earlier, a choke does not change total reactance (XL) in a linear manner vs frequency because its DC resistance affects the total reactance at low frequencies more than at high frequencies.
Conversely, a capacitor's reactance, on the other hand, is near zero ohms at high frequencies and rises to only a few dozen ohms at 20hz. A capacitor does have ESR so it is not perfect either. A decoupling capacitor does not perfectly isolate the choke from the signal circuit, especially at low frequencies, where the choke is most nonlinear. So the power supply has a changing reactive component vs frequency.
- A choke has lower and high frequency limits. A non-inductive resistor is virtually, totally linear from DC well into the hundreds of kilohertz (khz).
- A choke has self resonant and ringing issues. At what frequencies depends on the core material, size, and winding technique, associated capacitance etc.
- Chokes "pick up" stray magnetic and electrical fields, so proper placement and shielding is required to minimize the problem.
- When used in low current or small signal situations, like grid chokes of small tubes, the hysteresis curve is the most non-linear, creating the most distortion. The choke is directly in the signal path.
- When large signals are present, such as driving the grid of output tubes or the plate load of a tube, core satuation, increases harmonic distortion, eliptical loadline problems becomes an issue. Distortion rises as the frequency is lowered. High distortion, low value coupling capacitors may lead to "False bass".
From the Radiotron Designers Handbook.
"(A) When two or more input frequencies are applied to a non-linear amplifier the output will include the sum and difference frequencies located about each of the higher input frequencies. For example with input frequencies of 50 and 150 c/s, the output will include frequencies of 50, 100, 150, and 200 c/s. Even if the lowest frequency is very much attenuated by the amplifier, the sum and difference frequencies tend to create the acoustical impression of bass. With more than two input frequencies the effect is even greater, so that fairly high distortion has the effect of apparently accentuating the bass."
"(B) Owing to the peculiar properties of the ear, a single tone with harmonics may be amplified, the fundamental frequency may be completely suppressed, and yet the listener hears the missing fundamental."
"These two effects assist in producing "synthetic bass" when the natural bass is weak or entirely lacking. It should be emphasized that this is not the same as true bass, and does not consitute fidelity."
This is one reason that even 2nd harmonic distortion is objectionable. (And intermodulation distortion is approximately 3 times that of harmonic distortion.) This is exaggerated as the music becomes more complex.
So we see that chokes are anything but benign while adding to the cost. One actually pays more for inferior sound, but more "iron" makes for "good" marketing.
The Proper Size Power Transformer for Preamplifier Class A Operation.
What size power transformer is optimum for pure class A operation, say in a preamplifier? First, Class A means the power supply voltage remains constant, average current is constant as well. For a solid state rectifier type, the only requirement of a power transformer is supplying the necessary average current and staying cool. For tube rectification, a large power transformer may be necessary due to the higher internal resistance of the rectifier tube. However, with proper filtering stages, the power transformer should become a non factor to sonic quality. In fact, too large is actually detrimental due to the cost (which you pay). A large transformer may make for "good" marketing and profit, but I am about the sound.
If, for arguements sake, a larger transformer (heavy) is better for "regulation"; then using a tube rectifier with its high resistance actually defeats the very purpose of using a large transformer in the first place.
According to the RCA Tube Manual, a 5AR4 rectifier tube has 160/ 200 ohms impedance per plate at 450 volts/225ma and 500 volts/160ma, respectively. At currents in the low milliamps, such as preamplifiers, the tube's plate resistance is much higher. Even in power amplifiers, the high resistance defeats the very purpose of the large power transformer.
Amplifiers operated in class AB or B amplifier operation is another matter entirely. These classes of operation causes the average current to vary wildly, thus large power supply voltage changes occur unless a large, low resistance power transformer is incorporated for better voltage regulation. Unfortunately as mentioned above, rectifier tubes have high resistance which negates the large power transformers regulation ability. Swing chokes are often added to help regulate the voltage, but they add reactance and have their own problems. (See more concerning chokes below.)
As one can see, there are many considerations when choosing a power transformer for a preamplifier. A large power transformer means nothing, or is simply an expensive patch for other problems in the particular design the manufacturer used.
Rectifier Tube vs Solid State Rectifier
This is an interesting subject I had to deal with in my own designs. Do solid state (SS) rectifiers cause sonic problems? Proponents claim that solid state diodes produce noise while tube rectifiers soften the sound. Well, I can measure down to microvolts and have yet to measure any noise voltage from solid state diodes (with capacitor input filters). Filter stages are designed to reduce any possible noise from the AC line. Maybe they don't use enough stages.
Secondly, the proof is in the pudding. Virtually no one has yet been able to tell if my 11A is in the system or out, so the preamplifier is virtually perfect. So the real question is why are tube rectifier proponents having sonic problems? An improperly designed power supply is one answer. Doing a few equations or using a computer program does not tell the whole story as the models are not complete.
Now I do avoid solid state transistors/FETs in the direct signal path because they create their own sonic problems and of course are not isolated from the signal circuitry. If regulators, constant current sources, or Mu circuits are used, tubes are a must since they are in the direct signal path. Using a regulator with the decoupling capacitor up stream places the regulator in the direct signal path. (Do a thevenin equivalent circuit for proof.)
What we want is what sounds the best, not what is popular, or expensive, or is "good" marketing strategy.
Marketing Hype; Materials Do Not Matter.
It is interesting that in all areas, such as medical, mechanical engineering etc, parts, materials have flaws. Nothing is perfect by any means. Yet some preach that electronic parts are magically perfect, or at least so accurate our ears cannot detect the flaws. Besides harmonic and intermodulation distortion, there are other flaws and distortion producers, such as ESR, DA. Again we find insufficient computer models when it comes to parts. It also demonstrates the lack of basic research performed by some. It might be worth mentioning that college courses/text books only give the tools to research deeper. Text books and/or classes were never an all inclusive, totally in depth knowledge base. Consider large corporations that have in depth research departments.
I like the bandwidth as wide as possible. For instance, if a preamplifier is -1db at 20hz, there will be a problem. At 40hz, -0,4db, at 80hz, -0,15db. However, too much bass also creates masking problems, masking and thus changing what one perceives.
Bandwidth also play an important role. That -1db at 20hz means a lack for several octaves. That is quite noticeable. However, a deep dip or resonance less than 1/3 octive will probably not be perceived at all. (Rane Corporation conclusion.)
If the specs are very good, the perceived perception can be either good or questionable. However, if the specs are poor, the music will never be optimum. Oh it can sound good, but not optimum.
Well that is it for now. Just because a part or component is expensive, large, and/or weighty does not mean it is better. Just because one uses a computer program or mathematical equations (does look intelligent though) does not mean it is accurate, or distortion free.
In conclusion, I think one gets a glimpse how complex it is to produce a component.
Each page is designed to help educate the consumer when purchasing gear. From time to time, other articles and products will be added; so stay in touch.
I would like to thank God for inspiring the designs and parts used. It has been a real adventure, with His inspiration leading the way.
The articles covering theory require a special thanks to Walter G. Jung for discussions and his articles entitled "Picking Capacitors". Without these articles and discussions, these pages would be of lesser quality.
Thanks to Svetlana and JJ for allowing us to use their vacuum tube graphics on our site.
SAS Audio Labs
503 W. Jefferson St. Suite 2
Morton, Illinois USA 61550
Email to: sasaudio at omnilec.com"
*sas audio labs, SAS Audio Labs, SAS AUDIO LABS, and the SAS AUDIO LABS banner are trademarks of SAS Audio Labs."
*copyright©: 05-17-2008 Updated 2-12-2018. All contents of this page article (except Sovtek, JJ, and Svetlana Tubes) are copyrighted. Any and all designs and schematics, layouts of all our components, term "lead to lead wiring", "lead to lead connecting", "we make music come alive", "we make music truly come alive", and "the last watt is as important as the first watt" are copyrighted. All rights reserved. No portion of this article may be reproduced without written permission from Steve Sammet at SAS Audio Labs.