[bummrush]

Send one to hi fi critic and see what they come up with

[Roger A. Modjeski]

I will start with a disclaimer: NB (note well) that I am not a digital amplifier designer, that I have only a passing interest in them and I have no opinion about how they sound. If any class D amp designer cares to state how he has solved any of the problems below I welcome his information. I also welcome any class D amp designer to state his noise specs, frequency response graph, output impedance from 20-20 kHz, short-circuit protection and level of emitted radiation at the switching frequency.  Does anyone make on that will drive a 10 uF capacitor (big electrostat) at 20 Khz?

I was working on an unusual class D amp about 19 years ago with Mac Turner, a very intelligent EE. Power MOSFETs were just coming online. The Motorola first edition data book that I have is dated 1984 and is about 1/2 inch thick. By the time Mac and I were at it the IR book was about 2 inches thick and MOSFETS were cheap and easy to get.

The earliest commercial Hi Fi class D amplifier I know of was the Infinity SWAMP which I saw at CES in 1975 driving their SERVOSTATS. Kudos to them for trying and putting a lot of money at it but they were just too soon, the fast switching MOSFETs we have now were not available. They used the fastest switching transistors available but they were not nearly as fast as what we have now.

I know there are many terms for these amps and many designs floating around now; B&O, Tripath and others. I would not call these "digital amps" though I see that term used now and them. They are actually closer to analog amps until you get to the output stage and that's a BIG, FAST Switch. Keep in mind that the switching needs to be very fast just to get the amp to go to 20 KHZ. Fast switching also simplifies the output filter, which is the biggest, and not often mentioned, problem in these amps. Also note that these amps are not necessarily "faster" than other amps and often do not have even as wide bandwith as many tube amps notwithstanding their often derided output transformer.

Now here's a real bugaboo. Many of these amplifiers have feedback before one gets to the output filter thus the filter is out of the loop. What happens then is the amp becomes sensitive to load impedance and will either peak up if the load is higher impedance than the design load or roll off if it is lower. I do recall seeing a test of such an amp in Stereophile in the past few years and it did just that.

[Roger A. Modjeski]

Well, some of what you say is true ... but not always the case with tube gear. Generalizing will lead to religious battles ... let's not go there ... however I must say that...

Roger Modjeski's designs at Music Reference aim to be:
1) Good value for the dollar
2) Environmentally friendly and efficient
3) Reliable ... both in the short and particularly over the long-term
4) Sound good

I think Roger has specifically focused on these areas because he's been around the block, and at one point in his career repaired many a tube amp of other brands and makes, and saw exactly the downsides that you say. He tends to be very no-nonsense, no baloney type of guy (engineer) and has a track record to prove it with some very desirable gear floating around as part of his legacy. He's well respected (the RM-200 is Stereophile Class A rated) in audio circles though his positions on things hi-fi related also has created detractors.

One of the reasons I've bought and use MR gear is because I like this no B.S. philosophy, and in particular, his willingness to speak openly about his design implementations and design objectives.  Which is why his perspectives on some of the truths and myths I listed above I think are insightful to anyone interested in hi-fi.

I'd be curious as to your thoughts on the above list of 10 points?

Thanks 6BQ5 for stating my design principles simply and accurately. The RM-10 that I brought out in the early 1990's was the first tube amp where I gave great attention to power consumption. The best measure I could come up with was: rated watts out/idle watts in. This is what is happening up to the point where you play the amp really loud (near clipping) where line power does go up but is dependent on program material. In most listening situations the line draw moves up less than 10% from the idle value. The RM-10 came in with a rating of 70watts RMS capability with 60 watts idle, a number a bit better than 1 (1.16). A Dynaco Stereo 35 which has the same output tubes draws 85 watts from the wall at idle making its green factor 35/83 = 0.4, very close to 3 times more line draw for half the output power. Now these are small numbers but I obtain similar numbers in the RM-200. Even the little EM7-5 does well for a class A single ended amplifier drawing about 40 watts for 10 watts of output. It plays efficient speakers loud enough for my taste and draws less than an RM-10 in absolute terms. We know there are lots of 8 watt (16 w. total) 300B amps out there that draw more than 50 watts and easily up to 100.

[Roger A. Modjeski]

Tom Evans actually sells a tube power amplifier, though this preamp is SS.
There is more information on his Lithos regulator and preamp design here.
http://6moons.com/audioreviews/tomevans2/vibe.html



Quote:
"Tom claims that commercially available voltage regulators suffer voltage noise equivalent to the output of a moving coil cartridge. If you want to uncover all of the information contained in very small audio signals -- like those output by moving coil phono cartridges -- you want an audio signal with the minimal possible corruption and maximum possible dynamic range. This requires vanishingly low noise. If noise and signal are similar in magnitude, a significant amount of data is irretrievably lost. Off-the-shelf regulators also exhibit very slow transient response and recovery times, compounding the signal degradation begun with the noise issue. Tom decided that the only alternative was to develop his own high-performance high-speed regulators, which he christened Lithos. His first efforts at designing them yielded results that were a staggering 1000 times quieter, 53 times faster and 100,000 times more accurate than the best commercially available regulators used for audio applications at the time."

Here too, is another after-market regulator, though, no astounding claims made by Allen Wright.
http://www.vacuumstate.com/index.tpl?rubrik=13&lang=2




Both of these caught my attention, because (in relation to the 10 truths and myths), its only recently that I've seen any emphasis on power supply regulation as having much of an effect on sound in a preamp or amplifier. For the longest time, power supplies seemed to be all about beefier power supply caps.

I took a good look at Tom Evan's website and found the following spec on his Linear A amplifier.

---LINEAR A SPECIFICATIONS---
Frequency Response: 12Hz to 90kHz-these are 0dB down points! Flat.
Power: 25.2wpc Class A
Output Noise: 150dB down 700microvolts(!)(not millivolts)
Output Impedance: 0.5 Ohm
Input Sensitivity: .7millivolts

Price: $8900

His noise output is not referenced so I'll give him the most generous spec. and compare it to his 25 watt output. The answer is 86 dB (20 log 14volts/.7 mv= 86 db), not 150 dB. This makes me doubt his power supply noise claims also.

The Frequency Response spec makes little sense  "0 dB down" how about 1 or 3 like the rest of us do.

My output noise is typically 150-300 uV, yes microvolts, not millivolts.

The Input Sensitivity is likely .7 volts not millivolts.

Hey, I've had some errors in my specs too, none of us is perfect. I hope he will clean this up especially since he is claiming noise numbers that are not achieveable with active regulators unless you have a resistor feeding a big cap filter after them which both slows them down and degrades the output impedance.

[AmpDesigner333]

Tom Evans actually sells a tube power amplifier, though this preamp is SS.
There is more information on his Lithos regulator and preamp design here.
http://6moons.com/audioreviews/tomevans2/vibe.html



Quote:
"Tom claims that commercially available voltage regulators suffer voltage noise equivalent to the output of a moving coil cartridge. If you want to uncover all of the information contained in very small audio signals -- like those output by moving coil phono cartridges -- you want an audio signal with the minimal possible corruption and maximum possible dynamic range. This requires vanishingly low noise. If noise and signal are similar in magnitude, a significant amount of data is irretrievably lost. Off-the-shelf regulators also exhibit very slow transient response and recovery times, compounding the signal degradation begun with the noise issue. Tom decided that the only alternative was to develop his own high-performance high-speed regulators, which he christened Lithos. His first efforts at designing them yielded results that were a staggering 1000 times quieter, 53 times faster and 100,000 times more accurate than the best commercially available regulators used for audio applications at the time."

Here too, is another after-market regulator, though, no astounding claims made by Allen Wright.
http://www.vacuumstate.com/index.tpl?rubrik=13&lang=2




Both of these caught my attention, because (in relation to the 10 truths and myths), its only recently that I've seen any emphasis on power supply regulation as having much of an effect on sound in a preamp or amplifier. For the longest time, power supplies seemed to be all about beefier power supply caps.

I took a good look at Tom Evan's website and found the following spec on his Linear A amplifier.

---LINEAR A SPECIFICATIONS---
Frequency Response: 12Hz to 90kHz-these are 0dB down points! Flat.
Power: 25.2wpc Class A
Output Noise: 150dB down 700microvolts(!)(not millivolts)
Output Impedance: 0.5 Ohm
Input Sensitivity: .7millivolts

Price: $8900

His noise output is not referenced so I'll give him the most generous spec. and compare it to his 25 watt output. The answer is 86 dB (20 log 14volts/.7 mv= 86 db), not 150 dB. This makes me doubt his power supply noise claims also.

The Frequency Response spec makes little sense  "0 dB down" how about 1 or 3 like the rest of us do.

My output noise is typically 150-300 uV, yes microvolts, not millivolts.

The Input Sensitivity is likely .7 volts not millivolts.

Hey, I've had some errors in my specs too, none of us is perfect. I hope he will clean this up especially since he is claiming noise numbers that are not achieveable with active regulators unless you have a resistor feeding a big cap filter after them which both slows them down and degrades the output impedance.

Roger,

First, let me say that it is an honor to be corresponding with you.

I posted some comments about tube amps in general here:
http://www.audiocircle.com/index.php?topic=58526.msg522734#msg522734

Also, let me comment that your designs are well beyonw typical from what I understand, and I wish to do an A/B comparison of my amps to the RM-200.  I thought a neighbor of mine had these, but he doesn't (I asked today).  He just bought a pair of Quad amps (KT-66?) and has several other tube amps that I have listened to very carefully.  Thanks for your open discussion, and I do plan to go through the original list tonight.

Best Regards,
Tommy / DAC

[AmpDesigner333]

I will start with a disclaimer: NB (note well) that I am not a digital amplifier designer, that I have only a passing interest in them and I have no opinion about how they sound. If any class D amp designer cares to state how he has solved any of the problems below I welcome his information. I also welcome any class D amp designer to state his noise specs, frequency response graph, output impedance from 20-20 kHz, short-circuit protection and level of emitted radiation at the switching frequency.  Does anyone make on that will drive a 10 uF capacitor (big electrostat) at 20 Khz?

I was working on an unusual class D amp about 19 years ago with Mac Turner, a very intelligent EE. Power MOSFETs were just coming online. The Motorola first edition data book that I have is dated 1984 and is about 1/2 inch thick. By the time Mac and I were at it the IR book was about 2 inches thick and MOSFETS were cheap and easy to get.

The earliest commercial Hi Fi class D amplifier I know of was the Infinity SWAMP which I saw at CES in 1975 driving their SERVOSTATS. Kudos to them for trying and putting a lot of money at it but they were just too soon, the fast switching MOSFETs we have now were not available. They used the fastest switching transistors available but they were not nearly as fast as what we have now.

I know there are many terms for these amps and many designs floating around now; B&O, Tripath and others. I would not call these "digital amps" though I see that term used now and them. They are actually closer to analog amps until you get to the output stage and that's a BIG, FAST Switch. Keep in mind that the switching needs to be very fast just to get the amp to go to 20 KHZ. Fast switching also simplifies the output filter, which is the biggest, and not often mentioned, problem in these amps. Also note that these amps are not necessarily "faster" than other amps and often do not have even as wide bandwith as many tube amps notwithstanding their often derided output transformer.

Now here's a real bugaboo. Many of these amplifiers have feedback before one gets to the output filter thus the filter is out of the loop. What happens then is the amp becomes sensitive to load impedance and will either peak up if the load is higher impedance than the design load or roll off if it is lower. I do recall seeing a test of such an amp in Stereophile in the past few years and it did just that.

Roger,

Sorry about the bad spelling in my last post.  My laptop battery was about to go, and I wanted to post before I had to start over.

You are correct about the history here.  Making a good digital amp was nearly impossible back in the 70s and 80s.  I tried, though, with lack-luster results.

Regarding output filter designs, I could write a book on this.  Wrapping feedback around the filter actually has some serious downside.  I have designed amps like this and the compensation required (even using various modulation techniques) can defeat the purpose.  If the output filter is designed just right, it is only affected by the load at ultrasonic frequencies.  Even then, the difference is only a few dB from open circuit to 4 ohms.  Take a look at the effect on output amplitude here (DAC4800A/Cherry):
http://www.digitalamp.com/DAC4800A%20standard%20measurements%20v3.pdf
http://www.digitalamp.com/cherry%20measurements%20v6.pdf

Several years ago, I went to the trouble of plotting the compound response of the output filter driving the actual speaker load for a few different speakers, and it came out very flat.  Many speakers also have controlled impedance in the crossover that makes sure the natural rise in tweeter impedance is kept in check and doesn't do to open at ultrasonic frequencies.  The problem I've seen is that many "digital amps" (Class-D, really) are not such good designs, and come from engineers with very little experience in the field.  Going from analog amp design to digital amp design is not easy, and many manufacturers just wanted to get something to market to show that "they can do it too".  I've been at it for 20 years, and this is the only reason I can be confident in my conclusions.  However, the proof is in the pudding --- listening.  Thanks again, and have a great day!

Warm Regards,
Tommy / DAC

[6BQ5]

I hope he will clean this up especially since he is claiming noise numbers that are not achievable with active regulators unless you have a resistor feeding a big cap filter after them which both slows them down and degrades the output impedance.

Obviously something is amiss. I'm surprised no one else caught this type of error earlier.

Assuming these regulators were actually substantially "better" than typical PS regulation (whatever that may mean given the dubious specs) ... would this translate into a noticeable audible improvement?

[AmpDesigner333]

The story goes that audio-legend Paul Klipsch wore a button under his lapel that said Bullsxxt. He would flash it to someone who would engage with him on some audio-related topic (when appropriate of course).

I'm not at all an engineer, but like to read and learn to educate myself the gear I use and how they work (Roger's product manuals are great that way).

Over the years, I've come across various statements frequently made on discussion boards regarding amplifier design. As with all things, there is a bit of truth and some B.S. with each in the list below. These are NOT my ideas. Curious as to what others think ...

1.   Transformers: The bigger “iron”, the better, hence massive power supplies and output transfers sound best.

2.   Rectification: Tube rectification is better than diode, preferably with a choke in PS

3.   Tubes: NOS is generally better than today's Chinese, Russian, Czech and former Yugo production. It's more than just testing well.

4.   Parts I - Caps : Good ol’ Sprague caps sound better than new expensive exotics

5.   Parts II - Materials: Silver is better if you can afford it. Silver transformers, wiring, etc…

6.   Design: Fewer stages the better

7.   Negative Feedback: Preferably none. Zero NFB is best

8.   Design II: Class A … nuff said, generally sounds better than AB, B and other variants

9.   Construction: Hand wired is generally better than PCB

10.   Straight wire with gain : Tone controls are generally bad, another complexity to degrade the signal from "purity"

11.  Transformer II: Transformer-coupled amps perform better than those using a cap between stages

12.  Made in the USA: The Chinese don't care about quality of products. Made in the USA is best if you want your amp to last.

I would say with all of these, its a matter of implementation, thus "generalized" statements are mostly untrue. However, what is the "real story" with these? They can't be entirely right or wrong either.

I addressed each point on your list here:
http://www.audiocircle.com/index.php?topic=58526.msg522867#msg522867

This was put on my thread since I want my readers to see it.  Hope you don't mind.  Feel free to copy it back here.  There are posts on that thread with a LOT more detail about the same topics, so please read through and comment.  I'm a HUGE fan of good amp design, not just Class-D or Class-A or tubes.  I'm NOT a fan of gimmicks, and feel they are an insult to audio enthusiasts!

Please note that these answers are no replacement for a long roundtable conversation!  Thanks for your patience in addressing this list.

[Roger A. Modjeski]

Bad regulators can cause problems worse than simple R C power supply filtering. However, good regulators will give better results as they will have constant impedance over the audio band. In my studies of HV regulators in many popular preamps I was rather disappointed with their performance. In addition, most of them are not short-circuit proof which I find unforgivable.

[6BQ5]

Many of the items discussed in this thread are covered in this excellent new article by Nelson Pass on the Six Moons website - in particular: negative feedback, amplifier classes, and distortion measurements. Very much worth a read and quite an education on amplification!

Nelson Pass http://www.sixmoons.com/industryfeatures/distortion/distortion.html