[Freo-1]

Roger, my feeble attempt was in the previous post.  When I got the A2-120, I noticed that many of my "go to recordings" suddenly had what seemed to be additional information presented in a clear and concise manner.  I would put back in a couple other SS amps, and sure enough, it was missing (and a slightly noticeable) added sizzle to the treble.  Cymbals had more zzzzzz, as opposed to a natural ring. 

So, no, many designers did not (entirely) get the message.  The thing is, most folks would not notice it unless they hear an amp that does it correctly.  It took me a long time to realize most SS amps fall short.

[G Georgopoulos]

Roger, my feeble attempt was in the previous post.  When I got the A2-120, I noticed that many of my "go to recordings" suddenly had what seemed to be additional information presented in a clear and concise manner.  I would put back in a couple other SS amps, and sure enough, it was missing (and a slightly noticeable) added sizzle to the treble.  Cymbals had more zzzzzz, as opposed to a natural ring. 

So, no, many designers did not (entirely) get the message.  The thing is, most folks would not notice it unless they hear an amp that does it correctly.  It took me a long time to realize most SS amps fall short.

Are you comparing tube to ss?,yes tubes have zzzz as opposed to natural ring...

It took me a long time to realize most SS amps fall short.

not short, different!..

[G Georgopoulos]

Does anyone here want to take a shot at explaining slew rate and TIM in simple terms? The papers cited so far are written for engineers and not the layman. I have enjoyed reading them, but they are pretty heavy stuff for the general public.

Hi Roger

slew rate is the time taken for the rise or fall of a waveform,fast slew rate means less and less time taken for rise and fall..

cheers

[Roger A. Modjeski]

Hi Roger

slew rate is the time taken for the rise or fall of a waveform,fast slew rate means less and less time taken for rise and fall..

cheers

That is the definition of rise time which is taken at small signal levels of a few volts. Slew rate is how fast the amp can go rail to rail. To test for this one puts in a high frequency sine or square wave large enough to run the amp at full level. At some high frequency the wave becomes triangular and the slew rate is the slope of the rise (and fall) of voltage/time expressed in volts per microsecond. It literally measures how fast the output of an amplifier can change. An amplifier can have a one microsecond rise time at low levels but may not be able to rise that fast at higher levels such as 50 volts which is the peak output of a 100 watt amplifier.

Transient Intermodulation Distortion (TIM) occurs when the amp receives a signal that is rising too rapidly for it to follow. While it is trying to catch up no other information can get through thus the intermodulation. Again early amps like the Phase Linear 400 and 700 did not take this into account and had rather poor slew rates. Matti Otala and Electrocompainet were the first to bring this to the attention of the public and to designers who had not considered the problem.

It is well known that TIM was most often caused by high frequency transients (pops and clicks) from phono cartridges. When Matti Otalt did his research slew rates of some popular amplifiers was very poor. Look at the dates of those papers, around early 1980s. I think it is safe to say that digital sources will not produce signals that are so demanding and our concern about slew rate might be re-examined.

One needs to be careful making this measurement to disconnect any Zobel networks across the output of the amp as they will usually go up in smoke during the test. Bryston makes amps with good slew rates. Looking at theirs I find numbers like 60 V/us. Full power bandwidth is another way to get an idea of slew rate and theirs is full power to 100 KHz. I would say that any amplifier that can do full power to beyond 20 KHZ is adequate from a slew rate point of view.

For comparisons sake I found the specs on the Phase Linear 400 MK II (1977). The slew rate is 18 v/us and the rise time is 4 us. Even at this lower slew rate the amp almost makes it to the rail in 4 us. However slew rate is rail to rail and this is not quite fast enough to make that. The amp still had crossover distortion due to the output transistors being off as previously mentioned.

[G Georgopoulos]

Roger,slew rates and tim make ss amp unstable,how do you rectify that...

[Roger A. Modjeski]

Roger,slew rates and tim make ss amp unstable,how do you rectify that...

I don't know about that specifically. I will say that depending on how someone goes about improving slew rate may result in an amp that is less stable. There are a lot of amps out there that are unstable (go into oscillation) with small capacitive loads like 0.1 uF. I had a big Adcom on my bench years ago. When I put a 0.1 cap on the output the power line watt meter pegged at 1000 watts. I quickly removed the cap load.

There are many SS amps that are stable into the standard 2 uF load that will be unstable into smaller cap loads. I thinks some designers (engineers) assume that if the amp is stable into 2 uF it will be stable into anything smaller, which is not the case. This affects the end user in curious ways. Some speaker cables have capacitance around that 0.1 uf area and will cause the amp to go up in smoke. We had a Levinson in our store do just that.

The balance between stability and good high frequency performance (not just slew rate but low distortion and good damping at high frequencies)is not an easy one to achieve. Although this is often touched upon in an EE education, if one is fortunate enough to find an analog design course,  it is done in a strange mathematical way. While the professor and text will go on about phase shift and Nyquist conditions for stability they really don't get into how to measure it and how do deal with the consequences in a practical way. When i was at UVA in 1973 we touched on this. I had already designed a few SS amps in my teen years (for my band) and I was ready for them to pour the knowledge into me. Unfortunately none of them had built any SS amps.

Forgive my digression: here is the problem we now face and will get worse. Little by little EE programs have become CE (computer engineering) programs and there are very few schools that still give a hoot about analog at all. This is one reason for me starting a school to teach these things, many of which I have had to learn on my own. As I have often said "Where is the next generation of Audio Engineers"

[Pneumonic]

Plenty of good discussion.

Here's my take on slew rates and TIM.

The argument is that large amounts of NFB results in slew rate limiting which ends up in increased amounts of TIM. So, if NFB is lowered we wouldn’t see slew limiting nor increased levels of TIM.

Those who argue against such a claim suggest that it is not the level of feedback that imparts slew rate limiting and the resultant TIM but, rather, it is improperly implemented circuits done with excessive compensation that is the real cause. 

AFAIK, there is no peer reviewed evidence that audible levels of TIM exists in properly conceived gear.

As for slew rate. Slew rate indicates the slope (steepness) of the wave the amp can produce and is just a handy way of combining amplifier power (voltage) and frequency response into a single, convenient, value. It’s specified as a change of voltage over a specific period of time. This single value is not really needed as you can get more meaningful info by looking at the amps power (voltage) and its high frequency response limit. But we like convenience it seems ….. even at the cost of inaccuracies. 

By example. For an amp to produce 15v at 1 Hz, its output voltage would need to slew at a rate of 15v/second. So, a very shallow slope if viewed on a scope.  By contrast, if the voltage was increased tenfold to 150v, the slew rate would need to be 10 times faster or, in this case, 150v/second. And the slope would be 10 times steeper. If the frequency is raised from 1 Hz to, say, 10kHz, the slew rate would rise by a factor of 10k and we now get 1,500,000v/sec (or, since # is getting very large and unmanageable, expressed as v/uS instead). For an amp that outputs 150v and has a linear high frequency response of 100kHz (as is the case with most SS amps these days), the slew rate would be 150v/uS.
 
We know that output voltage determine power so lower powered amps will have smaller slew rates than higher powered amps … at the same frequency. Also, amps that have lower frequency response will have lower slew rates than amps that have higher frequency response. But, we also have seen that an amp that is spec’d out with a very high frequency response (ie 100kHz) will result in a large slew rate. But, us humans can’t hear above 20kHz so, in this case, the amp that goes to 100kHz won’t sound any different but, for a given power, it will have a slew rate that is 5x higher than that of the amp that is rated for 20kHz. And, you can almost bet that it’ll be deceitfully marketed as such. Thus the inaccuracy noted above.

It all can be confusing. Just note that, while slew rate must be high enough, a high slew rate, in and of itself alone, might not define the sound of an amp. Since all quality amps linearly produce 20kHz, it really is the voltage (power) that makes the difference when it comes to slew rates.

[Roger A. Modjeski]

Plenty of good discussion.

Here's my take on slew rates and TIM.

The argument is that large amounts of NFB results in slew rate limiting which ends up in increased amounts of TIM. So, if NFB is lowered we wouldn’t see slew limiting nor increased levels of TIM.

Those who argue against such a claim suggest that it is not the level of feedback that imparts slew rate limiting and the resultant TIM but, rather, it is improperly implemented circuits done with excessive compensation that is the real cause. 

I would say the first statement is untrue and the second is true. Lets look at how this happens and  how the first statement is related to the second statement.

Consider this: Without changing anything just disconnect the feedback on a SS amp. The ability of it to go from rail to rail (the slew rate) is unchanged. In my thinking this is proof enough that the first statement is untrue. The slew rate of an amplifier is not dependent on feedback but what you might do to apply a lot of feedback.

The slew rate of an amplifier is determined in the driver stage where a small capacitor (the compensation) is applied to make the feedback roll off at 6 dB/octave rather than something steeper that is happening from intrinsic capacitance in the devices and circuitry. This is called "dominant pole compensation" There are lots of ways to do this, some better than others. To optimize this takes some time on the bench.

Since op amps were done the easy way and it's in all the books, many engineers didn't give a lot of thought to what happens when the output of the amp cannot follow the input signal. Then you have TIM. However if one studies op amp application books and really understands them the information on the better ways to apply compensation are in there. There is very little difference between what is going on in an op amp and what is going on in a power amp. The important difference is that the guys who design op amps really know their stuff. RJ Widler, the founder of National Semiconductors is a brilliant guy. National is not likely to put out an op-amp that is flawed or miss-specified. They knew all this stuff 40 years ago when they brought out the 741 which still is still popular to this day (not for audio, but for many instrumentation applications}. I have used them in a power supply I developed just last year. I actually had to buy them and was happy to see they are still made while many other opamps have been discontinued. I will leave it up to the imagination of the reader why many designers of amplifiers don't know this stuff. 

I bring up the 741 because it is a rather slow op amp which makes it very stable even with 100% feedback (unity gain). Perhaps this method of compensation is the only one some guys know. In some amps I have tested I wonder if the designer knows much about compensation at all, hence the statement in your quote "excessive compensation". If one is not clever one just keeps increasing that little capacitor till the amp appears to be stable on his bench.We hope he hooks it up to a speaker and checks what happens there. A woofer near resonance puts some heavy demands on stability.

In cases where the designer is not watching the High Frequency (HF) performance of his amplifier he will simply increase this capacitor till the amp becomes stable. If he adds more feedback he has to slow the amp down even more thus inviting TIM.

While I want to defend the value of an engineering education I know that some engineers do not look too deeply into a particular design. This may be because of pressure from management, laziness or a lack of knowing where to look. People without an engineering education need a stronger flashlight and may miss many things. Of course experience can be a good substitute for an engineering education but some guidance may be required. There is nothing like finding someone who is doing good work in a field and studying with them.

I think most of you know that in some cases the designer is learning his skill while making amplifiers he does not fully understand. I have seen many instances of this. In most cases these designers and companies disappear. In other cases the designer learns how to do things better and better as time goes on.

[G Georgopoulos]

Roger,too short of a life too much to learn,your statement,thanks for sharing that with us...

[G Georgopoulos]

I think the problem is nfb,phase shifts can not be rectified if you have a global nfb loop,I think without global feedback and just local you could expand frequency response and slew rates as well,but then you have higher distortion which is what i dont want,so i see it all the time with global nfb making amps unstable if you try to run them faster and faster.

[Roger A. Modjeski]

I think the problem is nfb,phase shifts can not be rectified if you have a global nfb loop,I think without global feedback and just local you could expand frequency response and slew rates as well,but then you have higher distortion which is what i dont want,so i see it all the time with global nfb making amps unstable if you try to run them faster and faster.

If the amplifier, before feedback, has the proper bandwidth and 90 degrees phase shift when the feedback goes to zero there is no problem with global feedback. However one must consider what the load does to modify that situation.  So in some cases it might be better to have the feedback around the output stage separated from the driver stage. This has been done in several amplifiers.

In my new OTL the feedback is global because the open loop satisfies the criterion for stability. The path is extremely short. One gain tube, one phase inverter (no gain) two output tubes in totem-pole transconducting voltage to current and no push-pull transformer to have to deal with.  But more than that the way the feedback is applied is entirely different. In that amp it wouldn't work to go local. Every design is different and largely unto itself. That's what makes this fun.

[JoshK]

By the way, really great thread here!   Roger you are a great teacher.  I am in total agreement with your earlier posts on the explanation of class A, etc, and really started to learn a lot from the subsequent posts.

To date, I've only ever built class A (tube) amps.  That is all I have studied and since I find that I rarely need more than ~30wpc for all my listening habits, I never had need to learn how to design class AB amps for more power.   However as I've learned a great deal about speaker designs and crossover theory, I am increasingly suspicious of low dampening factor amps, so the idea surrounding feedback or parrallel output devices becomes more interesting.

[Freo-1]

By the way, really great thread here!   Roger you are a great teacher.  I am in total agreement with your earlier posts on the explanation of class A, etc, and really started to learn a lot from the subsequent posts.

To date, I've only ever built class A (tube) amps.  That is all I have studied and since I find that I rarely need more than ~30wpc for all my listening habits, I never had need to learn how to design class AB amps for more power.   However as I've learned a great deal about speaker designs and crossover theory, I am increasingly suspicious of low dampening factor amps, so the idea surrounding feedback or parrallel output devices becomes more interesting.

I heartily agree.  Roger has helped refine my understanding of several areas (Class A, TIM, Slew Rate distortion).  What's really hard is trying verbalize how issues like TIM and slew rate affect/modify the sound.  You "know it when you hear it", but explaining it in words is not very easy.  With today's SS amps, it's more subtle than it was in the past, but most SS amps still have some issues with the clarity regarding low level reproduction.  There are a VERY FEW that get it right, but not many (IMHO).

[Roger A. Modjeski]

I heartily agree.  Roger has helped refine my understanding of several areas (Class A, TIM, Slew Rate distortion).  What's really hard is trying verbalize how issues like TIM and slew rate affect/modify the sound.  You "know it when you hear it", but explaining it in words is not very easy.  With today's SS amps, it's more subtle than it was in the past, but most SS amps still have some issues with the clarity regarding low level reproduction.  There are a VERY FEW that get it right, but not many (IMHO).

SS amps that have trouble with low level detail are likely not slewing, it must be something else. Slewing is high level phenomenon. Here are some potential reasons for loss of low level detail: One thing people do love about tubes is low level detail. It is interesting that some manufacturers of SS amps boast that their midrange is as good as a tube amp.

1. Crossover distortion due to low bias (I wouldn't expect any modern amp to have this problem)

2. Too many devices in the signal path. A SS amps has several gain stages each with several transistors. Per channel the transistor count in a simple amp is about 10 minimum for the driver plus one pair of output transistors. A tube amp can be done with a one or two stage driver and a pair of output tubes or a single output tube in SE amps. Some SS amps have 20-50 transistors in the driver stage. When it comes down to it transistors are inexpensive, easy to put on a circuit board so designers often come up with exotic circuits that may look better on paper but not sound as good as simple ones.

3. Thermal problems. Transistors unlike tubes are very temperature sensitive.
 
4. Distortion due to high currents in wiring. Tom Holman discovered this when at Advent. He found that magnetic interference became the limiting factor in reducing distortion. Tube amps have high current only in the wires from the secondary to the speaker terminals.

5. Just too many parts overall. SS amps are rarely simple in comparison to tubes. Every time you add transistors you add some resistors and capacitors to run them.

6. Tubes are inherently more linear devices as we have said before.

[AJinFLA]

SS amps that have trouble with low level detail are likely not slewing, it must be something else.

I wonder if it has anything to do with the "test" method being used? Hmmm.....

Oh, to answer your earlier question regarding TIM/slew in 2014, my answer would be, what TIM/slew in 2014???
Perhaps a concern for the vintage folks?
Also, as Cordell pointed out, there were many (competent) designers not having any such issues before Otalas "breakthrough" papers.

cheers,

AJ

[AJinFLA]

5. Just too many parts overall. SS amps are rarely simple in comparison to tubes. Every time you add transistors you add some resistors and capacitors to run them.

Yeah, can you imagine how absolutely destroyed and degraded the signal/sound must be, passing through thousands of resistors and caps?
Like at a studio, including multiple passes through huge mixing boards, mic preamps, etc. with el cheapo commodity parts, permanently imbedding that massive degradation onto the storage playback media.
Playback media that "audiophiles" will then use, to easily and clearly "hear" a single resistor or cap in the (amp).

cheers,

AJ

[Freo-1]

Yeah, can you imagine how absolutely destroyed and degraded the signal/sound must be, passing through thousands of resistors and caps?
Like at a studio, including multiple passes through huge mixing boards, mic preamps, etc. with el cheapo commodity parts, permanently imbedding that massive degradation onto the storage playback media.
Playback media that "audiophiles" will then use, to easily and clearly "hear" a single resistor or cap in the (amp).

cheers,

AJ

Sadly, thanks to the "loudness wars"  much recorded modern music does indeed sound like crap. 

[bummrush]

Coda

[AJinFLA]

Sadly, thanks to the "loudness wars"  much recorded modern music does indeed sound like crap.

Classical, Jazz etc. have been largely immune, but yes, much popular music is horrible. Which matters scant to me if I enjoy the music.
But when it comes to system evaluation...

cheers,

AJ

[Roger A. Modjeski]

Yeah, can you imagine how absolutely destroyed and degraded the signal/sound must be, passing through thousands of resistors and caps?
Like at a studio, including multiple passes through huge mixing boards, mic preamps, etc. with el cheapo commodity parts, permanently imbedding that massive degradation onto the storage playback media.
Playback media that "audiophiles" will then use, to easily and clearly "hear" a single resistor or cap in the (amp).

cheers,

AJ

If you are saying that all those parts in the studio harm the sound then I agree. I am working on a tube reproduction (tape playback) amp to mate with the ubiquitous Studer A-80 "Master Tape Recorder" In looking at the schematics for this popular machine that was used for the 24 track initial recording and 2 track master tape that went to the record company, I am amazed at how many components the signal must go through. The play preamp must have 20 transistors and 100 other components. After it goes through all that it has to go through a line amp (to drive the 600 ohm line to +24 dBm). That line amp is a little power amp with as many components as before. For convenience of set up, the play level is adjusted in two places (two volume controls). There is an input transformer to step up the head output 10 to 1 and an output transformer to drive the 600 ohm line. Of course all the cards plug in and there are connectors all over the place. The gain is so excessive that the main repro volume control sits very low meaning the signal is amplifier to much higher levels then that gain is thrown away. Its a wonder it sounds decent at all.

There is renewed interest in playing master tape copies in the home on such machines. re THE TAPE PROJECT with electronics by Bottlehead. With my repro preamp I can replace the entire signal chain described above with just 3 tube stages which employ no more than 20 associated parts all-together. There are no transformers in the signal path and with a high impedance head I have noise a bit lower with my tubes than with their transistors. It does indeed sound better.