Improve a Rotel amp THD by 20dB!

Member
Joined 2012
Paid Member
Question: How good can an old Rotel get? I have had dozens of Rotel amps through the repair workshop over the years, and this RA-802AX came in as a “speaker killer”. It turned out to be fully working – except it had a solid +16Vdc on both Left and Right speaker outputs. In a Rotel, the 12A rated power transistors and the 4A output “protection” fuses will easily deliver the resulting constant 2Adc into 8ohm speakers (even more into 4ohm), or 32W heat dissipation which indeed can destroy most static standing woofer coils. Finding the unusual fault involved a thorough circuit analysis – and in turn led us to discuss Rotel's power amp philosophy which has changed surprisingly little over the years. Then came the wager. I put on a bet that I could easily modify the Rotel amp design to improve the Total Harmonic Distortion (THD) from the 0.03% in the specs by a decade – (i.e. a factor of 10 or -20dB in audio terms) – transforming this “entry level” amp into an audiophile class of less than 0.003% THD. We even added three tough rules: 1. Fully reversible changes, i.e. no cutting of PCB tracks. 2. Budget: max GBP 40, i.e. faulty components could be replaced, but no upgrades to any fancy audiophile high-end components. 3. Preserve the “Rotel sound” i.e. no major changes of power amp audio signal path topology. In short, it was down to basic good audio engineering and circuit analysis. Ho Hum. We agreed that the changes should be staged and at first only applied to the Left channel, so we could follow the improvements through comparison RTA distortion measurements. If anyone is interested in whether such a dramatic reduction in Total Harmonic Distortion can be achieved, I will describe each stage and the THD improvements in subsequent posts.
 
  • Like
Reactions: 1 user
Member
Joined 2012
Paid Member
RA-820AX mod stage 1

Ok, please do comment or ask questions if this becomes too boring or if there are things that needs extra explaining. After a succesful repair of the (very unusual) original +16Vdc output fault, the DC offset level on the speaker terminals were down to 60-90mV, as commonly found on these amps. While now clearly not destructive (
 
sounds interesting ... do tell.

but before you get started, just so i understand, please comment on " fully working – except ... +16Vdc on both ... outputs"
are you saying it had 16Vdc superimposed on top of the usual audio output signal?
what did you fix to cure it and what do you think caused it?

mlloyd1
 
Member
Joined 2012
Paid Member
Ok mlloyd1, since my original post attempts seem to have gone haywire anyway, I'll start by replying to your question.
Yes, this was something I'd never seen before. All the amp transistors seemed to measure out ok, you could even set the bias - and send a small audio signal through the amp - but there was a solid +16Vdc on BOTH outputs, frying innocent woofers - and almost also my test resistor setup. It did take quite some circuit analysis time, but eventually it turned out that both the base resistors (R611/612) in the input current source were o/c! Say what? Completely surrounded as they are by much more fragile semiconductors - they would absolutely be the very last component you'd expect to fail in the case of e.g. some surge/overload - or whatever. I have no explanation, but as these are the only two 6K8 metal film resistors in the amp, my only guess is that perhaps Rotel may have got a faulty 6K8 resistor lot?
 
wow, that's interesting.
about 20 years ago, i fixed a friend's old pioneer receiver that had that exact same fault, but only on one channel. the resistor had a hairline fracture that caused an open circuit. however, for this amp, the affected power amp channel appeared to have no power supply at all, so the output had no voltage of any kind, even the output devices were cold. it was like someone had pulled all power fuses.

i think adcom even used that as an approach to protectively shut down a power amp in case of a fault.
:)


Ok mlloyd1, since my original post attempts seem to have gone haywire anyway, I'll start by replying to your question....
 
Typical 3 stage gain "Sansui type" High Ft classic amp.

What is weird , they use a phase inverted driver stage (diamond like)
and then use positive feedback.
(below)

The NFB version of this amp (sansui) . is typically 20ppm-20k.
That is , with faster 35mhz sanken outputs.

OS
 

Attachments

  • Rotel.jpg
    Rotel.jpg
    103.4 KB · Views: 4,935
  • Like
Reactions: 1 user
Member
Joined 2012
Paid Member
mlloyd1, may I first say that I find it fascinating to meet someone who also finds such issues interesting. Normally I only get polite smiles and rolling eyes from those around me when I rant about these things.

I should add that after replacing the two 6K81's, the DC levels came back to normal and the amp seemed to play ok. However, I always do a final check for HF stability and on one channel there were small repeated bursts of oscillation - of very low amplitude and easily missed. Reason? My guess is that it was probably a damaged p-n junction in one of the input transistors (Q603/4) that was switching between going capacitive and conducting? After all, it had seen about 10Vdc of reverse Vbe and thus about 0.5mA reverse current from the feedback network - for who knows how long.

That spelled bad news for the repair, as both input pairs were now suspect and needed replacement with hFE matched pnp's. Further, the feedback DC blocks (C605/606) would have been stressed, as they are only rated 6.3V.
So a relatively small repair (albeit a tricky one) expanded - and led to the discussions and THD reduction wager which I hope to get on with as soon as I get my posting skills up to scratch.
 
Member
Joined 2012
Paid Member
RA-820AX mod stage 1 - once again

After a succesful repair of the original +16Vdc output fault, the DC offset level on the speaker terminals were down to 60-90mV, as not uncommonly found on these amps. While now clearly not destructive (<1mW/8ohm), such offset levels are not acceptable in a high end amp.

1st mod stage

So, first stage was to find a way to bring this down to single digit mV levels as there are no offset trim options in this amp. The offset is mainly caused by imbalance in the DC resistive values to the bases of the input transistors, both in the pre-amp IC and the power stage.

The preamp IC's in this 820AX were trusted old NE5532's, which I swapped to better OPA2134 audio opamps with low Ib JFET input stages, reducing this stage offset to less than 1mV.

I could then remove all electrolytic capacitors between the two stages, directly DC coupling these and, with a bit of resistor value calculations, establish a near perfect balanced DC load on the power amp input while retaining the same attenuation levels through both the source direct and the tone circuits.

I also established a better AC decoupling in the current source for the input transistor pair.

Stage 1 results:

Speaker output terminal offset: Lch (mod): 1mV vs. Rch (non-mod): still 82mV

THD level: down by an impressive -10dB to 0.0103% ….....Half way there! And already better than e.g. the NAD's!

THD+N (distortion+noise) was also reduced – but from a terrible 0.7649% to a still too high 0.3387%, mostly coming from 50Hz line noise and its harmonics. (Please note that this measurement should be taken with a pinch of salt, as our RTA setup seemed to pick up 50Hz noise during a direct tape-out loopback test – even with the amp switched off.) But the relative improvement is valid.

I enclose a spectrum analysis showing the distortion peak improvements (blue to red) and the direct loopback in grey. Note the 50Hz and odd harmonic peaks on this "reference" trace - but no even harmonics (e.g. 100Hz).
 

Attachments

  • ra-820ax org (blue), 1st mod (red), tape out ref (l. grey).jpg
    ra-820ax org (blue), 1st mod (red), tape out ref (l. grey).jpg
    103.9 KB · Views: 3,385
Member
Joined 2012
Paid Member
Interesting. I currently have an old RB960BX on the kitchen table and have been considering some tweaks. My first thought was to add a current mirror to the input stage. It will be interesting to follow your progress.


Yes, a current mirror in place of the single R605/606 collector resistors would greatly improve things - but was deemed "not allowed" as "being in the power amp audio path topology".
Sic, talk about moving goalposts!
 
  • Like
Reactions: 1 user
at east electronics the book says that we have repaired 73 Rotel of that line the past 5 years .

All of them had cleaning issues...
From 73 there was 32 of them that had failure enough to make them not work ( beyond cleaning issues )
From these 32 the 25 had this type of failure no more than a TO92 transistor with a minor leak , resistors open , and a notorious Miller cap shorted ...

So i would say that for Rotel that doesn't look uncommon.

Kind regards
Sakis
 
Member
Joined 2012
Paid Member
at east electronics the book says that we have repaired 73 Rotel of that line the past 5 years .

All of them had cleaning issues...
From 73 there was 32 of them that had failure enough to make them not work ( beyond cleaning issues )
From these 32 the 25 had this type of failure no more than a TO92 transistor with a minor leak , resistors open , and a notorious Miller cap shorted ...

So i would say that for Rotel that doesn't look uncommon.

Kind regards
Sakis

Hi Sakis,
Thank you for your interesting stats.

Yes, I agree that these amps can be proper dust collectors, although this one was actually not too bad.

In my experience, the most common fault in the Rotels is due to people shorting the outputs, at best blowing the fuses or often the faster "Three-legged fuses", i.e. the output transistors.

BTW, which Miller cap have you found as "notorious"?
 
Hi Sakis,
Thank you for your interesting stats.

Yes, I agree that these amps can be proper dust collectors, although this one was actually not too bad.

In my experience, the most common fault in the Rotels is due to people shorting the outputs, at best blowing the fuses or often the faster "Three-legged fuses", i.e. the output transistors.

BTW, which Miller cap have you found as "notorious"?


Another possible cause of failure although I don't know how common, is through someone increasing the Iq. In that regard r641 and r643 the 910R resistors from the driver emitters are not crossed coupled to each other but to the opposite voltage rail. There is a recent instance of failure due to excess heating, of driver stage components.
 
Member
Joined 2012
Paid Member
I forgot to add that as part of the mod I consider adding emitter degeneration resistors to the input pair. Simulations seem to indicate that it would work, but I am a novice.

The best advice I received many years ago from my uni professor was to "always do things in stages and learn from each step - or you will find yourself having to start over and over".

Yes, emitter gm degenerating resistors can give improvements (if closely matched and the long tail pair current properly increased, etc.), but there are much simpler mods to be done first to an old amp that can dramatically improve performance - as I hope you may eventually see from my planned mods.