Improve a Rotel amp THD by 20dB!

Disabled Account
Joined 2015
I'm not sure if it's OK to tweak the circuit itself. The "sound" usually has been tweaked back and forth by experienced constructors. They have made small alterations on biasing, topology, etc until they find a sound they think people like and that is in line with the company's audio ideal. Fiddling with these things may just turn the amp into harshness, that's the usual result, hitting the sweet spot isn't so easy. I think it is like that.
But improving the "power plant" such as caps seems to just make things better, preserving the "sound".
One exception from that. I had a NAD preamp that started to sound bright and overly clear when I substituted the transformer to a toroid one. Before it had a cuddly, a bit english sound.
Another NAD amp I had was split up between preamp and power amp. I found that the power amp sounded clinical and bright, and the preamp like it had a wool coat drawn over it - very soft. Probably the engineers tried to compensate the bright power amp with a soft preamp. But that's not a good thing, you shouldn't compensate one flawed part by another flawed part in the other direction.
Ordinary "blameless" amp category with a lot of NFB seems to be suited for tailoring sound. In contrary to what most people believe, NFB actually gives an amp it's coloration, it doesn't remove it.
There is a hifi guru - Cordell, I think his name is, argues just that. The NFB is a very strong coloring factor.
Personally I like the esoterical stuff with low NFB, class A etc.. It's like scotch malt whisky, the taste isn't "taylored". It's the result of choice of barley, water, etc...
 
Administrator
Joined 2004
Paid Member
Hi Svitjod,
Feedback only does it's best with what it has to work with. The "sound" of an audio product is given by the circuit types and choice of input stage components. Then there are the parts choices in key locations. All these things create the "sound" of the audio component. Feedback can modify the character of an amplifier (say), but it doesn't create the type of sound.

It's all cold, hard science from steel hearted engineers working in a secret lab. Instruments their only guide.

Feedback is attributed for many things. It isn't something that is bad. Feedback is rather good, but has it's limits with slow circuitry where the phase angle shifts too far in the amp. Then you have to be careful how much you apply. But generally speaking, the more feedback you can apply to an amplifier that is "blameless", the better off you are. That is the honest truth of the matter.

-Chris
 
Disabled Account
Joined 2015
Anatech, you seem very sure when you speak about these matters. We shouldn't be too sure when it comes to subjective matters such as hifi. The great thing about NFB is actually the amount of antagonistic feelings it provokes among people. :)
I just cited the hifi guru Cordell. He thinks NFB has a very coloring effect on the subjective impression.
I think the subject is controversal because the concept is so damned elegant and appeals to every educated engineer. Still subjectivists grumbles.
My own humble opinion is as follows. ( I'm going to try to be short )

An amp has normally about 4 or 5 stages which all distorts a bit. When global NFB is applied the error signal rattles through the system in a complex manner. The effect is that even if the THD gets lower, the complexity of it rises. Furthermore, the higher the frequency gets, the harder it is for the feedback to correct properly. Then we have the matter of stability. Hard NFB lowers the phase margin.
This is only my own reflections which I can't prove in any way, but I have found that an amp benefits not only from a "safe" margin of lets say 45 degrees, but will perform better subjectively if the margin is very large - that is; "rock stable".
This may be one explanation why low NFB amps may sound better.

OK, you perhaps think I'm nuts with my talk about "complex" distortion. Isn' every distortion bad?
For example, 2:nd order dist is very hard to notice, even several percent. But the much lower cross over distortion that comes from a class AB BJT amp is more annoying, since it is a real burst of overtones. When these artefacts rattles around in the NFB loop things gets even messier.

So, either a low amount of NFB. Or a very short feedback loop that goes around perhaps only two stages, such as in Nelsons minimalistic amps.

OK, don't be too harsh on me, I'm just trying to illuminate the subject, I may be totally wrong.
 
Member
Joined 2012
Paid Member
I sometimes have a suspicion that we are so used to bright sounding amps that we actually think this brightness is a sign of clarity. Personally, I prefer amps that makes you relax deeper in the sofa until you become one with the music. Some sort of zen experience, and then we don't wan't overly bright amps.
Don't you agree?

Hejsan Svitjod,

Who am I to disagree with any Zen experience? :angel:

I am not quite sure that I can follow your explanations about NFB and THD, except that yes, a distorting amp can indeed sound "harsh" or maybe "bright".

But there is definitely a big difference between the audible properties of "bright" vs. "clear" (or perhaps rather "detailed"). Just like you, I used to love the soft Rotel sound, but will never go back again. I have been asked this before, see posts 74 and 76 and 168 almost exactly one year ago:

So, the Rotel RA-820AX project is finished – and just in time for Christmas. I will put a “To Me from Myself” sticker on it and put it under the tree. And then (with due copyright credit to Chris' wife's bon-mot) I'll go “OH NO, you shouldn't have!” when the unwrapping time comes.

The original goal was to improve the THD by -20dB on a GBP 40 budget.

It ended up -26dB (or 20 times better than stock) – for a total of about GBP 60 due to the minor mains transformer mishap.

Now, to all those who – perhaps quite rightfully - claim that THD is really no measure of sonic or acoustic quality, could I maybe recommend the book: “Loudspeakers for Music Recording and Reproduction” by Phillip Newell and Keith Holland. It is one of the most insightful books I have read in a long time on the issue of subjective and objective acoustical assessment. It puts forward the idea that THD is “the top of the iceberg” of what we hear as “good” or “bad” sound reproduction.

And, if you bring down THD by reducing amp non-linearity (which is what I have done), you will also have greatly reduced the other gremlins (in particular IMD). I quote: “An increase in input signal which produces a negligible effect on low-order [harmonic] products can wake up the `evil forces` of nonlinearity, releasing an unfathomable number of high-order intermodulation product `piranhas` to tear the flesh of the reproduced sound to pieces”. Well, I couldn't have put it better myself.....

Anyway, “Well, How Does It Sound?” I hear you cry. As I have stated previously, I am really not very good at producing Hi-Fi press wooly superlatives, so let me try to perhaps give an analogy to how I perceive the “sound” of a non-modded Rotel amp to what I have now:

“Go out on a clear night in the suburbs of a city and look at the starry sky, and you may say: “Oh that's nice, look how many stars you can see!”. And then – with a few modded steps - go out in a dark field or a mountain desert and look up – and you will go “WOW!”

That's how it feels.

Merry Cristmas everybody!
 
Member
Joined 2012
Paid Member
Svitjod
Indeed, that is the goal but it seems to be a moving target for me :)

AngelP or Per if you prefer, this is one of the most eye opening and entertaining threads on diyaudio. It's a real pleasure to see techs with real world experience troubleshoot right before our eyes. I have one question about projects like these:
How do you know which amps are not likely to be worth tweaking?

Thank you for the nice comments, nania.

Amps not worth tweaking.....hmmm let's see what springs to mind:

  • High end amps - already great
  • Very low end amps - not worth it
  • Digital or integrated amps - how, why?
  • Cosmetically damaged or simply ugly amps - yuk!
  • Amps with heavily corroded pots and connectors - too much work
  • Classical amps - will lose value if modded
I am sure that others can/will add to (or disagree with) this list.

Cheers,
Per
 
Member
Joined 2012
Paid Member
Hi Angelp,
i have a rotel ra-970bx. in complex music passage it is muddy and in midrange it lacks . i have the schematic. can you suggest the mods to improve the sound quality?

Hi Suresh

It is very difficult to diagnose and/or suggest improvements over the web, not knowing what the real problem could be, the level of electronic competence you are at and what test equipment you have at hand.
It is all too easy to do severe damage to the amp, the speakers (or indeed to yourself) by small, silly mod attempt mistakes. Been there, done all that (but please don't tell anyone I said so;)).

The first thing you could do, however, is to measure the DC level on the speaker terminals (should be much less than 60-80mV with min volume setting). If not - you'll need a repair before any mod. Then, I would suggest getting in touch with jpc2001 for pics and advice as he is currently doing exactly the same work as you look for on his RA-970. I have never modded this model.

Cheers,
Per
 
my amp doesnt have any problem and offset is well below 30mv. i have multimeter and oscilloscope and i am an electronic engineer. all i want to try and improve the sound . i was expecting some mod on compensation schemes. ra-970bx uses shunt compensation instead of miller . by removing shunt and placing miller will help improve the sound quality?
 
anatech said:
Feedback is rather good, but has it's limits with slow circuitry where the phase angle shifts too far in the amp.
This is the most concise and accurate explanation of using feedback I have ever heard. +1
The only thing I can add is that the where the feedback is applied will dictate the "color" Svitjod is referring to. I also agree with Svitjod's assertion that adding sequential stages of signal processing makes the sound less natural. For me, there is some lushness and realism that is lost when you go above 4 sequences. There may be tricks to prevent this over-processing sonic phenomena but I am unaware of them. The legendary John Curl wrote about this too and prefers fewer stages in his designs. In my opinion, it's unwise to ignore his advice.

AngelP
Thank you for your attention and reply. I draw the "high end" amp line at 50ppm single tone THD+N and three tone IMD distortion at full power. You'll be surprised how many "high end" amps fall short of these specs. We are about at the anniversity of your remarkable effort. Your mid-fi Rotel has become a truly impressive performer and I can't help but wonder if some classic vintage Pioneer and Kenwood gear couldn't be coaxed to a similar level of performance. Most notably the Pioneer dual mono integrated models (9500/8500-II) and the JFET Kenwoods. There are others of course but these are fairly ubiquitous and have huge potential for layout improvements because they're built on perfboards. Also, their looks still hold up even to this day. Has anyone done any mods on these and what is the consensus on their potential?
 
Member
Joined 2012
Paid Member
Hi Suresh,
Thank you for clarifying that, it seems that we are on the same level. I hope you appreciate the need for caution on advising on these matters.

But as I said, I have never modded a 970 and just looking at the schematics some investigation of the real amp is needed. It seems that there are shorts between the Q619/621 driver bases to the rails - that simply can't be so, nothing would work.

Anyway, take it step by step - for a quick first mod I would suggest putting in 100 E / 100uF 63V rail filters as jpc did and then measure/listen. Then consider the ground star issues and the reservoir caps. Here again the schematic says 8,200uF / 50V - if this is actually true - this is much too close for comfort with 49V rails and I would replace with 63V rated 10,000uF's.

This should firm things up and then we can look towards changes in the VAS configuration.

Happy modding!

Per
 

Attachments

  • Ra-970bx pic.jpg
    Ra-970bx pic.jpg
    259.3 KB · Views: 488
Administrator
Joined 2004
Paid Member
Hi nania,
Most circuits actually break down into similar circuit blocks. Added circuitry should normally be created to make the actual signal carrying elements more linear. The result is that the schematic may appear to be more complicated when it isn't really. If there are added stages, they would tend to be in the output stage to reduce non-linearity in that stage (current amplifier). Again, these added parts make it look more complicated than it really is.

I own amplifiers with both global negative feedback, and some that only employ local feedback. One extremely capable amplifier using that scheme would be the Cyrus Mono X and others in recent time. They sound pretty darned good, but then the development utilized an capable engineer, and listening tests. Since some of my own work was along this line, I understand it easily but to others it may seem to be too complicated. Of the standard type amplifiers, my favorite so far is my Marantz 300DC. A complicated looking thing, but then again, looks deceive. It's actually a pretty straight forward design that works extremely well. I have to admit that I've played with the circuitry myself, but the basic idea is the same.

Feedback only causes trouble when the basic circuit has design issues. With a good solid design with very low distortion to begin with has feedback added (you can't use them open loop), the existing distortion is reduced and the sound quality improved. If you take the amplifier out of its normal operating range, it will misbehave of course. Sometimes feedback can make the result worse, but you have to understand that once you saturate (or starve) a stage, the feedback loop is opened and you're now dealing with the full open loop gain without feedback employed, with predictable results.

-Chris
 
Administrator
Joined 2004
Paid Member
Hi Svitjod,
Well, I've designed amplifiers and modified quite a few as well. I don't listen to my own work, but there are some people who do. I just send them a piece and see what comments come back (I'm not even present for the listening tests). Their only instructions are that positive comments do me no good, tell me what is wrong. My findings over many, many years is that my measurements agree with the subjective listening tests. I look at the residuals from the distortion analyser. The changes in readings can be small compared to the subjective change, but once you know what you're looking at you can sometimes even predict what the subjective comments might be. Can't do this every time, but when I can, I'm on the money.

So it's all about experience and subjective feedback. From that you can sometimes get a pretty good read on something from measurements and just simply knowing what some circuit configurations might sound like.

-Chris
 
Disabled Account
Joined 2015
This is a nice thread, the subject is so wide that almost anything can be discussed.
Yes, modding is fun, I used to do that before. Nowadays I find it more stimulating to construct things from scratch.
But there is one pitfall when you start modifying things, and that is the famous placebo effect. I have noticed that every time I have made some alteration that makes the object sound different, one tends to think it's an improvement, just because it's something new and exciting - there is a new intersting flavor present.
This normally settles after a while. Sometimes after a couple of hours, sometimes after half a year.
I have noticed( on my own reactions ) that if I get overly enthusiastic about a mod, then I should be cautious.
On the other hand, if I notice a relaxed feeling or if I tend to be absorbed by the music, that is often a good sign. Nowadays I use to feel it in my legs if something is right, they feel kind of a bit heavy and tend to rest more reaxed at the floor. ( advanced placebo )

So, the thing about making improvements is that it's a slow process and we always have to fight against the placebo demon. But in the long run we tend to wind up with things we can live with.
That's probably why people often fails at blind tests. It's hard to get an opinion so quickly. And then, the perceived sound of a hifi gear is a combination of the actual sound and our own expectations. In blind test, we are mainly unsure, and we can't really use our feelings.
 
anatech
We agree that when it's applied correctly, feedback will always improve the circuit result but I don't think we can always know when it's applied correctly. I view feedback as current/voltage momentum. The trouble with momentum arises when you need to change direction or slow it down. A signal wave can be described as accelerating and decelerating and I have some theories related to this momentum description why current feedback is less imposing on circuits but that is another discussion. I might open it up in another thread but it doesn't belong in here. Group delays along with signal smearing/straining can be introduced with the application of feedback and hidden by great THD numbers. I compare this to video/graphic processing/sharpening where divots and hairline imperfections are either smoothed or shaved off. The result is cleaner and sharper but it is definitely changed. Few would call this change distortion, see my point? Generated signal tones are much simpler and cleaner than recorded music so just because it looks like it should when we measure it doesn't mean more complex waves will.

No one took up my earlier question about vintage gear. Here is one example:

https://www.manualslib.com/manual/892645/Pioneer-Sa-9500ii.html?page=28#manual

Is this a potential winner or no? If yes, I'll start a new thread on it.
 
Last edited:
Disabled Account
Joined 2015
Yes, nania, heavy NFB can be like the video sharpening algorithms on TV's. They appear to be sharper but are they really?
My personal reflection on heavy NFB is that it makes the music more strained, as if the musicians hade stage fright and were very ambitious to put every note right.

Vintage gear. That reminds me of the seventies, when high slew rate was very hot. Every amp should exhibit at least 50v/us or so.
 
Disabled Account
Joined 2015
Just one more thing. Indeed NFB is nessessary on a class AB BJT amplifier. The cross over artefacts must be lowered otherwise the amp will have a tiresome sound.
But using a large dose of it requires skill and experience, I think. And a lot of listening tests just to find one of those sweet spots.
 
Administrator
Joined 2004
Paid Member
Hi Svitjod,
That's why I don't listen to my own work. I have some listeners who never know what they are going to hear and have come to understand they shouldn't have any assumptions going into a test. One of my prototypes was out for over ten years. She died and now it's coming home to papa for more work. Her family loves the thing.

-Chris
 
Administrator
Joined 2004
Paid Member
Hi nania,
Actually, everyone becomes confused about one simple fact about feedback. The effect of electricity travels just under the speed of light. Therefore, for all intents and purposes feedback occurs instantaneously. The equilibrium for feedback is attained immediately and there is no effect like a mass involved. Since there is no time delay to speak of, and we are talking about really, really low frequencies even in the FM band, cause and effect is instant for audio frequencies. Hmmm, doesn't that throw a new light on the entire question of feedback? Even the slow pokes in the circuit, transistors, operate above 100 MHz (most of them do). That is such a short time interval that at worst case, the entire question of feedback arriving too late to fix a problem is put out to pasture. This is a non-issue and I thought that reasoning it out would be better than just making a statement that the question has no basis in fact (and this is also true). So in your analogy, consider feedback as massless and infinitely quick. That should make your mental exercise make more sense.

You know a circuit is low distortion before adding feedback by simply measuring the distortion and looking at the distortion analyser and spectrum. It takes a practiced eye (which makes it look like magic to an inexperienced person) to determine these things. Engineers do it every day.

Group delay, another bogeyman. As above, the effects are infinitely faster than you can imagine, which brings group delay numbers into the nsec or lower. Don't worry about stuff that has no meaning. Group delay is a problem with filters and digital circuits, but not with a wide band audio amplifier. Frequency response filtering creates group delay as part of the effect. You can't have a filter without some group delay. Nothing bad happens and the world didn't blow up. Next time you give a tone control a twist, try to figure out what problem(s) that group delay caused. (the answer is none)
Is this a potential winner or no?
No. I have one and it doesn't seem to be anyone's favorite. Mind you, it has some stiff competition. However, this is only my view and I know that others will share your view. So start a new thread on it.

In the 70's when I was selling these exact amplifiers, Pioneer was very aggressive with their advertising. It worked and the brand was very popular, but other brands in the same time period outperformed them easily. It is one of Pioneer's better amplifiers though. Funny thing. At the store where I worked, dealers had to sign up to move a certain number of Pioneer (and other brands) per year or face financial penalties. One year we were always out of stock of JVC and had Pioneer stacked to the roof (not lying or exaggerating). We asked why we couldn't have more JVC and were told that we were committed to Pioneer. My response was that people wanted the JVC product (sounded a lot better, Super-A was out). I just said we could go broke missing the sales while we stock-piled Pioneer that no one wanted. We started getting JVC in about a week and a half later. We did sell a pile of Pioneer car systems (Supertuner). That made our numbers.

Life can be so strange.

-Chris
 
Administrator
Joined 2004
Paid Member
Hi Svitjod,
heavy NFB can be like the video sharpening algorithms on TV's.
Careful about generalizations. Heavy feedback is used in every op amp circuit, and every note you hear has been through hundreds of them before you hear anything.
My personal reflection on heavy NFB is that it makes the music more strained,
That's fine as a personal opinion. It depends on what you have listened to. There are a lot of great sounding pieces out there with heavy feedback that sound open and airy while also being soft, drawing you into the music. You need to hear some better equipment I think. :)
That reminds me of the seventies, when high slew rate was very hot
You know, the average equipment reached a pinnacle in design around 1979 ~ 1983. Some stellar products came out scattered throughout time. 1959 for the Eico product and others like Fisher and McIntosh, 1970's you had the Marantz equipment (the 500 was way ahead of it's time, intro in 1969 until 1974). You had Sansui, Luxman and some other brands that made wonderful stuff. Even in the 2000's, Cyrus has been cutting edge and the reintroduced Marantz as well. But the bulk of development was in that critical time before everything fell apart and became a sales numbers game. Consumers have been ripped off by a poor market for quality now. Still, there are some great pieces out there. When I rebuild an older set and use modern components and circuit ideas, I end up with performance that far exceeds the original and also what is current. The circuit designs and construction was very good at the end of the 70's. Some slight modifications can really do wonders for that equipment.
Indeed NFB is nessessary on a class AB BJT amplifier. The cross over artefacts must be lowered otherwise the amp will have a tiresome sound.
Not really. I have some Cyrus Mono X amplifiers that do not use global feedback, just local. Really good stuff. As long as you bias an AB stage out of class B, you're fine. Ask Nelson Pass about feedback with his Stassis circuit that leave the outputs outside the feedback loop. Hmmm, oops! Looks like you don't actually need feedback on a class AB output section.
Be careful when making generalizations. I burn myself doing this as well.

-Chris
 
anatech
I kind of knew you would come back with that argument. You should know that the speed of electricity is limited by its conduit so it'll very rarely move near the speed of light unless the voltage remains high and the conduit is adequate. The operating speed of active devices act very much like the operating speed of a CPU, there are process lags and leakages everywhere within the silicon. These are what create the synchronization errors in the signal we call group delay. Once these distortions are introduced, they are buffered over and polished to make them look like there was no molestation. These would not be visible in a nice, clean, predictable, generated from a precision instrument tone. Music is not characterized this way, at least for me it isn't :)

I know I can't prove this yet so I completely understand why you might eschew my thoughts on the matter but ceteris parabus, I like feedback localized and performing a useful function (PSRR/CMRR) and I like fewer sequential stages. Given the noise floor is <-108dB, I also like a dominant harmonic profile (odd or even) better than I like equal amounts of H2+H3. May I ask what makes the SA-9500/9500-II a poor upgrade candidate for you? When I looked inside one, I remember noting the expensive stuff is adequate and the parts could be reseated on perfboard. Did you see the schematic I linked?
 
Last edited:
Disabled Account
Joined 2015
Interesting. Now I seem to have an opportunity to discuss these matters with an experienced constructor.
The effect of an NFB is immediate, that's true, but at higher frequencies the error signal has to fight itself forth through several unlinearities and capacitors wich all makes the output signal more messy. The problem, as I see it, with NFB is that the effect diminishet at high frequencies.
I think it was John Lindsey Hood that once said something like this.
"Applying NFB is a bit like hammering out a big dent in your car. The dent will largely be hammered out, but will be replaced by a bunch of small dents in it's place".
And then we have the problem with stability. Heavy NFB decreases the phase margin.
My very subjective experience is that amps tend to open up sonically with a really large margin. That is, more than what is needed to make the amp stable.

Anatech, you said that a AB biased BJT stage may be run without any NFB that reduces cross over dist. Thats odd. If so, why use feedback at all?
Recently I finished a class AB, BJT amp. First I followed my own hobbyhorse rule and applied not too much NFB, lets say 26DB or so. It sounded open and vivid but I got a slight headache after some listening. The treble had a bit of barb wire quality.
Finally I gave up and applied as much NFB as the circuit allowed. Around 50 db up to say, 7khz. Luckily my circuit is very dense, there are only two stages inside the feedback loop. It ended up well. Barb wire gone. Silky and nice.
Thank's to the simplicity of the circuit, the phase margin is excellent, never below 120 degrees regardless of any capacitive load.

So.... feedback works best with short loops. But even they must be kept within stability margins.

Regards