Improve a Rotel amp THD by 20dB!

Negative DC feedback absolutely does correct for DC offset as long as the diff pair is balanced and you have no issues in the voltage amp stage. Then you would expect the diff pair to be severely unbalanced. With a good diff pair, the base voltages should be darned close (a couple mV worst case). Some amps will oscillate if you probe directly. YOu can try putting 100R resistors in series with your meter leads and measure between the bases.
 
I get ~17 with 22k, ~10 with 12k, ~8 with 10k.

So your spice model uses a hFE of about 300 for the KSA992? The ones I got have a hFE of 400. So i think I'll put in two 10k for R611/613.

But I actually meant the temp sensitivity with a lower gain, all things equal it should help things?
Sorry Per, I misunderstood. Temperature sensitivity doesn't change much, the DC offset I get with 5º difference is:
22k -> 109 mV
12k -> 99 mV
10k -> 96 mV

And with the cap in series with the 6k8 resistor:
22k -> 12 mV
12k -> 11.7 mV
10k -> 11.6 mV

I don't know about the gain, my KSA992 model is based on measurements of a sample with beta = 387... How did you measure it? I just took the voltage difference between the collectors divided by the voltage difference between the bases, and that's what I get...
 
@cabirio , while you are simulating - can you feed 1kHz to the amp without it starting to oscillate?
I seem to be unable to get clean 1kHz out of it (in simulation).
No problem here, I attach the .tran simulation file for your perusal. Also a 20kHz square wave shows just a little overshoot. On the other hand I'm still working on the stability analysis, with the unusual T-feedback topology I haven't found a good place to insert the Tian probe and get a reasonable loop gain plot. I'll report back if I manage.

820BX3-thd.png


820BX3-20k.png
 

Attachments

I attach the .tran simulation file for your perusal.
Thanks, now it simulates ok. Do not know what was wrong in the earlier version.

Edit:
How can I plot the input and output signals in the same pane with different Y-scale factors/limits?
So that I would see the 200mV input signal and 15V output signal as an overlay having the 200mV signal e.g. 50% height (in the plot) of the 15V signal.

Edit2:
Multiplying the input signal (in the plot trace definition) by e.g. a factor of 40 makes the traces visually comparable but there is no correct Y axis for the input signal values.
 
Last edited:
@cabirio , while you are simulating - can you feed 1kHz to the amp without it starting to oscillate?

Actually, in real life, a detailed analysis with a scope seems to show very small remnants (?) of HF. As far as I can make out it is one 100MHz and two in the lower MHz range, but all down in the uV range. I guess that the many lead and lead-lag cap compensations are there for a reason.

(And you should of course be mindful of noise from any surrounding SMPS or LED light and intermodulations thereof. Although that would probably show up as spikes on the trace rather than oscillations).
 
Rotala RA-820BX3 (contd.)

After a few days of break from this project I think it is time to take stock at what has been achieved so far – and what further could/should be done.

I think that have kept the original Otala basic 4-stage circuit design pretty well intact.
Only inserted, deleted and replaced a handful or two of components here and there.

1724101124596.png


This has
a) eliminated the serious overheating that scorched the pcb,
b)
reduced the 3rd and 5th harmonic distortion significantly and
c) reduced the out of control output offset drifting to the point where a speaker protection relay won’t be clicking on and off wildly.

Not bad for a few changes?👍

Note that in addition to what has been described in previous posts, I put in 10Hz LP ripple reduction LC filters for the LTPs (162R and 100uF).
I also replaced the input 10uF C601 electrolytic with a 3u3 film cap and reversed the 15k and 6k8 resistors to keep the LF cutoff at about 3Hz.
And removed the silly 1000pF C609 parallel cap. All quite straightforward to do on this pcb.

However, there is still some offset drifting and the THD could possibly be reduced even more.

Yes, I really should run an RTA at this point, but my first objective was to have a good listen to an Otala amp design – so that is what I plan to do in the coming days, now that I am more assured that my Angelhorns speakers are relatively safe from DC frying :flame: by this amp.

Stay tuned...
 
Last edited:
  • Like
Reactions: anatech and madis64
Rotala RA-820BX3 (contd.)

Hello,
Sorry for the hiatus in this project, but sometimes life (read: work :smash:) comes in the way of having fun.

And then, when I set up the measurement gear again - something had happened.
Apparently, the measurement noise floor had moved up by more than 20dB ! And the whole thing looks very strange indeed.🤔

R820BX3 strange mesurement issue.jpg


I will have to look into what has happened here before moving on with the 820BX3, so please stay patient.
Per
 
Finally found the problem!

The blessed ALPS rotary-to-sliding switches that are open to corrosion/fouling from the atmosphere.

1726995158860.png


Over the years a black muck and silver oxide layer can form on the contact rails inside. You can't get contact cleaner spray in - and it wouldn't be able to remove the layer anyway. So the only way is to remove both outer and inner front panels and then very carefully unsolder the multipin switch and take it apart. And the latter is not for the feeble hearted or ten thumbed diy'er! The pins can break and the phenolic board is brittle, plus the motherboard's solder pads do not like too much heat stress from all the soldering.
Further, if things really do go awry, these switches have of course been discontinued so long that it is impossible to find new spares anymore.
If only ALPS had gold plated the rails during production.....

Muck (remember that some of these amps have suffered the time where smoking in the home was more or less normal) acts as an insulator, while silver oxide acts as a semiconductor. A recipe for inexplicable intermittent contact issues.

And that was indeed the issue for the strange RTA traces. A bit of input selector switch wiggling - and the trace returned to normal. Sometimes.

So, it is on to switch extractions, disassembly and cleaning.🙄
 
I've had to do this already a few times in my 820AX, admittedly over many years, but eventually the problem comes back. I've sometimes wondered if there's a reasonably easy/cheap way to gold-plate the contacts, but haven't really looked into it, anyone? I've already lost a couple of pins, very luckily not connected ones, but next time I might not be so lucky...
 
Takes usually an hour per switch, with a coffee break.

Yes, and a few chosen words when you take one apart and find it in pristine condition inside......:headbash:

There is definitely a limited amount of times that you can un- and re-solder these switches and keep the pcb connections intact.
So I don't do this per routine as there is a risk/benefit balance to be struck.

I have been considering buying a gold plating pen kit to make the work worthwhile and lasting regardless of the severity of fouling.
 
  • Like
Reactions: Smartdriver
a reasonably easy/cheap way to gold-plate the contacts, but haven't really looked into it, anyone?
I just ordered the bare minimum stuff from SPA Plating https://www.goldn.co.uk/product/gold-pen-plating-kit/
The kit is expensive, so I only selected the very basic things that is needed, planning to use my lab power supply etc.
But it still added up to just over £200.:gasp:💸
 
  • Thank You
Reactions: cabirio
Rotel RA-820BX3

SMACK
:faint:

That was the sound of my palm hitting the forehead when I re-visited the 'Rotala' project.

I remembered that I had put in a speaker protection board early on - only to remove it again when the relays started to click wildly.
Still don't know why they did that, but I had of course removed two shorting bars to the speaker terminals to get the protection circuit in between.

And a brief study of the pcb revealed that the output feedback tracks actually connected close to the terminal (ie. now isolated) side.
So, I had been trying to get a stable circuit from something equivalent to this:
1727945453663.jpeg

An open loop 'comparator'.
No wonder it was hyper sensitive to temperature changes, etc.
In fact it is pretty remarkable that we got the "output" stable to within a few hundred millivolts .... or actually at all !
A quick re-routing of the FB connections to before the protection - and heavenly peace fell on the Finish landscape. From "Rotala" to "Oh la la!". :deer:

The good news was that all the efforts that was put into the amp did their job:
All self-destructive temperatures were down, offset is close to zero, stable (!) and easily fine adjusted with the trimmer in almost center position and THD is now down to 0.021%.

Time for a celebration brew, a supine position and a good listening test.👍