I think with Naim it is easy to look at the circuit from a macro viewpoint and frown, since from an intellectual viewpoint a full complementary symmetry design (from input to output) must surely do a better job.
I can't say that I have done any comparisons with a Naim although I currently own one. However I was able to do that with a Cyrus 1 many years ago and I preferred the sound of the Cyrus. Both makers use all NPN transistor output stages.
At that stage I was a convert to full complementary symmetry and had a vested interest in this since the test subject was a magazine project I had built for myself and had thought highly of it. It is now a partly dismantled relic in my garage and I have no motivation to re-commission it.
I am aware that computer modelling of circuit layouts has been around for a good while now and Naim were early exponents of this approach. The upshot of this was attention to such things as earth points, trace width, general circuit layout etc, and having parts manufactured to their own specifications - The power output transistors is one example. I think Cyrus amplifiers were manufactured by a military standard capable facility.
In either case I think both Naim and Cyrus are highly interested in circuit behavior at a micro level and that includes the layout and even the materials and rigidity of the enclosure.
When I bought my Naim amplifier I had it on conditional loan for several weeks trial and I could return it if I didn't like it. It was always most definitely going back until a week before the trial was up - I absolutely hated it.
I think one of the inbuilt survival mechanisms we still have from the hunter gatherer days is to be acutely aware of the sounds in our environment and it takes a while to get used to changes. I found it easy to dismiss the sound of something that did not meet the sensory perception I had always been used to.
My DIY amplifiers don't grace our lounge, as they are too bulky. The Naim isn't.
Yes, it would be informative, but by now I suspect that the conclusion would be "don't use this amplifier - get another".
I did a bit of improvement when recalculating the tone control components, as the stock tone control ranges were pretty paltry - about +/- 3-4 dB, particularly in the treble. I got that to work to +/- 6dB and was rather happy with that, hoping that I didn't have to re-do it as you have to go through the bother to remove all the front panels to get to the solder side of this PCB.
Eventually, it turned out that there was probably a good reason for the limited treble gain - as running the amp with the new treble gain at full +6dB, after some (?) time the music started to sound strained and the heat sink had gotten quite hot.
HF oscillation - power off!
This was in a standard speaker setup sort of as you describe above. After the heat sink and my nerves had cooled down, I checked the amp and speakers - and then turned things on again with the treble in mid position. No problems.
So, I've now got a Rotel that I can order into oscillation with the twist of a knob. My first idea was that rather than documenting what load stability limits this old amp has or may never have had, we could try the various HF stabilisation techniques in some sequence to perhaps see which one is most effective?
I do not have a plan for this yet - any suggestions from you guys?
Hi sgrossklass,
Good suggestions. I will now redesign the resistor load and Xonar U7 into one box, so I can get better control over ground paths, shielding etc. Wish me luck, I will report back in due course.
Before launching into changes of HF stabilization you might do a visual "check of the roof" to see whether or not there is any source of leaks.
In that regard the 0R22 buffer resistor at the output stage acts on the negative feedback loop and does nothing to buffer this from the speaker cable or the speaker itself.
I gather from an early post by Mooly that this resistor is a wire-wound component, which has a small amount of residual inductance, and which in theory is capable of picking up radiated fields such as those generated around emitter resistors and transistor supply loops in through Class AB switching actions. The resistor is in the feedback path where any contamination picked up will be amplified.
There are ways to deal with these issues and you would need to check these have been executed in view of the greater sensitivity to them.
Looking at the lag compensation scheme used there seems to be a path through the 150p capacitor from the negative supply rail through the 10 pf capacitor to the base of the voltage amplifier transistor. There will be switching artefacts on the negative supply rail.
The circuit for the earlier RA820A shows the 150p capacitor near the voltage amplifier connection to the negative rail rather than adjacent to the PNP power output for the RA820AX. The question is whether this is a lag component or does it address a hole storage issue or a bit of both.
Hi mjona,
The Naim products I have worked on in the past even have layout problems (goofs actually). It was clear that they were quite poorly put together using circuit techniques that only work at low enough voltages, or they would smoke. Never mind the schematic, execution was rough! Your model may have been much better than what I've had to look at over here.
Cyrus is a polar opposite having inspired design and quite good build quality. The only thing I didn't like was the finish in the casting areas for heat sink contact. In fact, I saw the Mono X in development on schematic only and decided to own a pair right off the bat. Here they sit, the very first pair in North America. I am not disappointed. These little stinkers output an honest 150 wpc each! Cyrus turned to a full complimentary output stage after the Cyrus one. Have a look at what they did in the diff pair stage, you will be interested.
Having met the team at Cyrus, I can fully understand why they sound as good as they do. These folks are professionals who do have the equipment to back up their listening tests. Doing both keeps them on the right track.
-Chris
The Naim products I have worked on in the past even have layout problems (goofs actually). It was clear that they were quite poorly put together using circuit techniques that only work at low enough voltages, or they would smoke. Never mind the schematic, execution was rough! Your model may have been much better than what I've had to look at over here.
Cyrus is a polar opposite having inspired design and quite good build quality. The only thing I didn't like was the finish in the casting areas for heat sink contact. In fact, I saw the Mono X in development on schematic only and decided to own a pair right off the bat. Here they sit, the very first pair in North America. I am not disappointed. These little stinkers output an honest 150 wpc each! Cyrus turned to a full complimentary output stage after the Cyrus one. Have a look at what they did in the diff pair stage, you will be interested.
Having met the team at Cyrus, I can fully understand why they sound as good as they do. These folks are professionals who do have the equipment to back up their listening tests. Doing both keeps them on the right track.
-Chris
Last edited:
Hi AngelP ,
Given all the new gain, you may have to use an RC network between the collectors of Q601 and 603. Same for the other side of course!
Have you looked at the oscillation with an oscilloscope? The positive driver and output has no miller compensation, or any other type. The bases have a 330pF cap to ground, which is maybe a little heavy. They may have set phase margin empirically, then you improved performance which shifted everything again.
Another very obvious shortcoming is the absence of a compensation cap across R637, probably the very first place I would look. Do look at the oscillation first as there are clues in there. Drive it with a sine wave partially and increase the treble control until it breaks into oscillation just near the peaks, + or - or maybe even both peaks. If you can get it beginning to take off, add a low value cap across R637 to see if it makes things worse or better. 10 pF ought to show you a direction.
-Chris
Given all the new gain, you may have to use an RC network between the collectors of Q601 and 603. Same for the other side of course!
Have you looked at the oscillation with an oscilloscope? The positive driver and output has no miller compensation, or any other type. The bases have a 330pF cap to ground, which is maybe a little heavy. They may have set phase margin empirically, then you improved performance which shifted everything again.
Another very obvious shortcoming is the absence of a compensation cap across R637, probably the very first place I would look. Do look at the oscillation first as there are clues in there. Drive it with a sine wave partially and increase the treble control until it breaks into oscillation just near the peaks, + or - or maybe even both peaks. If you can get it beginning to take off, add a low value cap across R637 to see if it makes things worse or better. 10 pF ought to show you a direction.
-Chris
I have had information about the Cyrus 1 and 2 circuits and Mission Loudspeakers of that era in my silent keeping for many years and there is much to admire in what I have seen.
This equipment like Naim was made in New Zealand for a short time.
I knew the agent for Mission/Cyrus after his enterprise folded and shortly before he emigrated to England. I had a secure job and income and was not interested in setting up in the electronic repair business. There were a couple of Cyrus repair jobs the agent passed to me which I did for my personal edification.
My impression is Cyrus hired a top designer in the field of operational amplifiers at a time when design was becoming somewhat repetitive.
Hi mjona,
They are designed a little differently, aren't they?
It seems that all the Mission distributors folded around the same time. I did work for Mission here, but the warehouse guy was an idiot. He decided that he knew how to service product and starved me for parts. He kicked all the big Mission systems my way. Remember the ones that had some speakers built in and did auto-setup for the surround? Yeah, I got those and nothing else unless he couldn't fix it.
I think he kept all the lasers when they went down. The speaker drivers for sure.
-Chris
They are designed a little differently, aren't they?
It seems that all the Mission distributors folded around the same time. I did work for Mission here, but the warehouse guy was an idiot. He decided that he knew how to service product and starved me for parts. He kicked all the big Mission systems my way. Remember the ones that had some speakers built in and did auto-setup for the surround? Yeah, I got those and nothing else unless he couldn't fix it.
I think he kept all the lasers when they went down. The speaker drivers for sure.
-Chris
Hi mjona,
They are designed a little differently, aren't they?
-Chris
Any ideas who by?
@AngelP,
Just wondering about your tone control arrangement working as an "oscillation control" after your changes in the operating ranges..
19 times seems a lot of line voltage gain for a tone control and that may have been the factor in Rotel's design choice of a low operating boost and cut ranges. Your modification will have lifted the bar on that one.
If you are going to stick by your mods I recommend that you look for an op.amp with better unity gain specifications than OPA2134 which at 8MHz is lower than the 10MHz capacity of NE5532. You would need a device with enough headroom to cover your extra treble boost. Alternatively you could reset the treble control to original and reduce the line gain by 20%.
When it comes to low voltage supply rail decoupling it is always good practice to decouple each device near the supply pins. Here according to the circuit diagram the phono stage and tone controls have shared decoupling which is sited near the former. It is an economy move which may be possible with devices of the same type, but I don't like the chances with different types.
Just wondering about your tone control arrangement working as an "oscillation control" after your changes in the operating ranges..
19 times seems a lot of line voltage gain for a tone control and that may have been the factor in Rotel's design choice of a low operating boost and cut ranges. Your modification will have lifted the bar on that one.
If you are going to stick by your mods I recommend that you look for an op.amp with better unity gain specifications than OPA2134 which at 8MHz is lower than the 10MHz capacity of NE5532. You would need a device with enough headroom to cover your extra treble boost. Alternatively you could reset the treble control to original and reduce the line gain by 20%.
When it comes to low voltage supply rail decoupling it is always good practice to decouple each device near the supply pins. Here according to the circuit diagram the phono stage and tone controls have shared decoupling which is sited near the former. It is an economy move which may be possible with devices of the same type, but I don't like the chances with different types.
Hi mjona,
I met them in 2004, just before I had a rather bad accident. I can't remember names at all. He was about my age and was designing further along the same lines I was going. I knew what those amps would sound like before they were even built (I had built prototypes on my own before meeting him). They sounded just the same as I expected them to.
I spent the better part of a day with them and became very familiar with their methods for design. Basically, if something doesn't sound right, back the design goes. They rarely need rework for audible issues. The entire company is driven in this same way, unusual to see these days away from the design by ear crowd.
The assembly is very efficient, and each unit was tested by Audio Precision equipment. The same thing in the repair department. Each repair also has the Audio Precision gauntlet to run, and service has their own machines. All I'm saying is that this company uses very good test equipment (they have Agilent as well), and good enough means that the equipment betters the listed specifications even for repairs.
-Chris
I met them in 2004, just before I had a rather bad accident. I can't remember names at all. He was about my age and was designing further along the same lines I was going. I knew what those amps would sound like before they were even built (I had built prototypes on my own before meeting him). They sounded just the same as I expected them to.
I spent the better part of a day with them and became very familiar with their methods for design. Basically, if something doesn't sound right, back the design goes. They rarely need rework for audible issues. The entire company is driven in this same way, unusual to see these days away from the design by ear crowd.
The assembly is very efficient, and each unit was tested by Audio Precision equipment. The same thing in the repair department. Each repair also has the Audio Precision gauntlet to run, and service has their own machines. All I'm saying is that this company uses very good test equipment (they have Agilent as well), and good enough means that the equipment betters the listed specifications even for repairs.
-Chris
I don't believe that is true.
The designer of the Cyrus 1 and 2 amps is a man called Henry Azima. Very smart guy.
Hi mjona and Chris,
Sorry for the delay in reply, sometimes I have to sneak some work in between the fun.
All IC's have been decoupled properly - actually the PCB holes and places were already there, but apparently some beancounter decided to save a farthing or two.
Also, note that the 10pF Miller capacitors (C611/12) are NOT mounted in the UK versions of the amp?? Did they consider the average British customer to anyway be too unstable to notice? Any other explanations? Patent issues?
And yes, the simple lag compensation with its massive collector loading of the VAS Q607/8 is brutal and really no way to treat a nice 2SC1941.
I am very tempted to start to upgrade the VAS stage - but a) will it then still be a Rotel and where should you stop? b) I need to sort out the RTA setup first and c) I need a plan and a relatively easy setup to measure the efficacy of the various HF stability circuit ideas.
Cheers,
Per
@AngelP,
How much Iq are you running through the output transistors?
The Technical Manual specifies a measurement of 4 millivolts across a 0R22 emitter resistor in each channel when the amplifier heatsink is warm after running at rated power into 8 ohms for around 5 minutes at no specified frequency - 1KHz I expect which should be safe. That would give an Iq of 1.8 milliamps under extreme conditions.
To my mind this is just too hit and miss and it would be better to do the adjustment with monitoring by a distortion meter and repeat the exercise so you can accommodate varying ambient temperatures of your listening environment.
I do wonder if the low Iq is a measure against cross conduction in the output transistors due to marginal stability of the feedback loop.
How much Iq are you running through the output transistors?
The Technical Manual specifies a measurement of 4 millivolts across a 0R22 emitter resistor in each channel when the amplifier heatsink is warm after running at rated power into 8 ohms for around 5 minutes at no specified frequency - 1KHz I expect which should be safe. That would give an Iq of 1.8 milliamps under extreme conditions.
To my mind this is just too hit and miss and it would be better to do the adjustment with monitoring by a distortion meter and repeat the exercise so you can accommodate varying ambient temperatures of your listening environment.
I do wonder if the low Iq is a measure against cross conduction in the output transistors due to marginal stability of the feedback loop.
Last edited:
@AngelP,
Looking now at the stability compensation and related components C607, C609, C613 and R623 and corresponding parts in the other channel, it may be worth noting the mention of the short failure of the Miller cap in post 16 - albeit these are not fitted in this version of the RA820. These parts [the capacitors] are all ceramic with a 50 volt rating. Also you had suspicious failures with 6k8 resistors.
I note the thinking that the value of C607 at 330 pf is heavy. That is assuming it is intact -neither open circuit or leaky.
If open circuit the Vas loading will reduce at high frequency. That will increase the frequency at which phase shift starts - reducing your stability margin.
There may be surety about this if you simply replaced them all [resistors too if necessary] if only to establish a sound starting point before any prospective alterations to your compensation.
Looking now at the stability compensation and related components C607, C609, C613 and R623 and corresponding parts in the other channel, it may be worth noting the mention of the short failure of the Miller cap in post 16 - albeit these are not fitted in this version of the RA820. These parts [the capacitors] are all ceramic with a 50 volt rating. Also you had suspicious failures with 6k8 resistors.
I note the thinking that the value of C607 at 330 pf is heavy. That is assuming it is intact -neither open circuit or leaky.
If open circuit the Vas loading will reduce at high frequency. That will increase the frequency at which phase shift starts - reducing your stability margin.
There may be surety about this if you simply replaced them all [resistors too if necessary] if only to establish a sound starting point before any prospective alterations to your compensation.
Last edited:
Hi Per,
You know what? Sometimes trying to retain the sonic signature puts you directly in conflict with reliability. I do try to retain "the sound", and in other cases I work really hard to eliminate "that sound". One example of that is with Pioneer equipment when the customer likes the look, but wants high sound quality.
Just for giggles and grins, "fix" the VAS stage. You can always put everything back later. Decide if the change is worthwhile (probably is).
-Chris
You know what? Sometimes trying to retain the sonic signature puts you directly in conflict with reliability. I do try to retain "the sound", and in other cases I work really hard to eliminate "that sound". One example of that is with Pioneer equipment when the customer likes the look, but wants high sound quality.
Just for giggles and grins, "fix" the VAS stage. You can always put everything back later. Decide if the change is worthwhile (probably is).
-Chris
No, 4mV over 0.22 ohm gives a normal Iq of 18mA. The small trimmers are a b****r to adjust anyway, this could call for multiturns or perhaps a redesign.
This I think is where commercial constraints come into play, things such as heatsink size and PSU capacity.
18ma is way below an optimum sort of value for this type of stage. If you read the work done by Doug Self back in the 1990's it was found that specific values of quiescent current for specific emitter value resistor values seemed to give the best thd performance. For this configuration of output stage you are looking at around 105 milliamps give or take as somewhere in the optimal range.
One problem with the Rotel (as you discovered) is the poor shared grounding. Increasing the quiescent current increase the ripple voltage at idle which makes the perceived hum and noise much worse.