Rotel RA-820BX3 ‘Rotala’ (contd.)
I got another RA-820 BX3 to see whether the issues I had in ‘Rotala’ #1 were unique for that particular amp or if there should be some systemic faults in these ‘non-Curtis’ designs.
The first thing of notice is that this BX3 ‘Rotala’ #2 has a completely different pcb and even some circuitry differences – without any indicative model name change, not even a “Mk.II”.
Well, that became clearer after a bit of de-dusting:🌪️
After some repairs of faults (yes, I bought it “with some problems” on ebay for £25). I turned it on and it ran just as flaming hot as Rotala #1.
The VAS to about 90 ºC and the 2W 910R driver base resistors got up to a whopping 120 ºC!.🔥
I actually found that these two resistors in the left channel had burned out (!), so the driver transistors only acted as b-e diodes.
The output offsets were -680mV and -595mV (!!) – which could not be trimmed out, but at least they seemed to be stable!
So I decided to first keep the right channel original as reference, and do the required changes to tame the temperature and offsets in the left channel.
The 'bare minimum' upgrade went precisely as smooth as expected. It did take me several hours as I (over-)carefully did a full DC, AC, temp and scope test after each individual change. All temperatures now under 55 ºC and the offset stable at zero with the trimmer in the centre position. No instability whatsoever. Very limited output change to varying temperatures – i.e. everything completely ‘normal’ for a Rotel. THD down at 0.020%, mainly 3rd.
So, if you should be so (un-)fortunate to purchase or be gifted one of the following ‘Rotala’ amps:
RA-820BX3 (version #1 and #2)
RA-820BX4
RA-840BX3
RA-840BX4
RA-940BX
RA-960BX
(there could be more, perhaps someone in this thread can add to this list)?
- you will have to take action because you will have issues of unwanted high output offset and suicidal pcb overheating as these amps all use the exact same Rotel version of the old Otala design.
So, as an absolute minimum you should immediately heat up the old soldering iron and get 14 inexpensive resistors (all 1% MF 1/4W):
4x 270R; 4x 10k; 4x 1k and 2x 82R.
Then replace as follows:
The 270 and 10k’s will remove most of the offset by balancing the LTP (and significantly reduce THD in the process). Trimming to zero is now easy with VR603/4.
The 1k’s will reduce the VAS current and bring the VAS temperature down from over 90 ºC.
The simple 82R (replacing the two big 910Rs) eliminates the heat they produce while doing the same job.
And you could then also remove C613/4 and short out R685 – the LTP’s 2SA1016s can easily take the full rail voltage.
Then trim the bias up to 4mV to reduce crossover distortion – and ‘Robert is Your Father's Brother’ as they say.
👍
I got another RA-820 BX3 to see whether the issues I had in ‘Rotala’ #1 were unique for that particular amp or if there should be some systemic faults in these ‘non-Curtis’ designs.
The first thing of notice is that this BX3 ‘Rotala’ #2 has a completely different pcb and even some circuitry differences – without any indicative model name change, not even a “Mk.II”.
Well, that became clearer after a bit of de-dusting:🌪️
After some repairs of faults (yes, I bought it “with some problems” on ebay for £25). I turned it on and it ran just as flaming hot as Rotala #1.
The VAS to about 90 ºC and the 2W 910R driver base resistors got up to a whopping 120 ºC!.🔥
I actually found that these two resistors in the left channel had burned out (!), so the driver transistors only acted as b-e diodes.
The output offsets were -680mV and -595mV (!!) – which could not be trimmed out, but at least they seemed to be stable!
So I decided to first keep the right channel original as reference, and do the required changes to tame the temperature and offsets in the left channel.
The 'bare minimum' upgrade went precisely as smooth as expected. It did take me several hours as I (over-)carefully did a full DC, AC, temp and scope test after each individual change. All temperatures now under 55 ºC and the offset stable at zero with the trimmer in the centre position. No instability whatsoever. Very limited output change to varying temperatures – i.e. everything completely ‘normal’ for a Rotel. THD down at 0.020%, mainly 3rd.
So, if you should be so (un-)fortunate to purchase or be gifted one of the following ‘Rotala’ amps:
RA-820BX3 (version #1 and #2)
RA-820BX4
RA-840BX3
RA-840BX4
RA-940BX
RA-960BX
(there could be more, perhaps someone in this thread can add to this list)?
- you will have to take action because you will have issues of unwanted high output offset and suicidal pcb overheating as these amps all use the exact same Rotel version of the old Otala design.
So, as an absolute minimum you should immediately heat up the old soldering iron and get 14 inexpensive resistors (all 1% MF 1/4W):
4x 270R; 4x 10k; 4x 1k and 2x 82R.
Then replace as follows:
The 270 and 10k’s will remove most of the offset by balancing the LTP (and significantly reduce THD in the process). Trimming to zero is now easy with VR603/4.
The 1k’s will reduce the VAS current and bring the VAS temperature down from over 90 ºC.
The simple 82R (replacing the two big 910Rs) eliminates the heat they produce while doing the same job.
And you could then also remove C613/4 and short out R685 – the LTP’s 2SA1016s can easily take the full rail voltage.
Then trim the bias up to 4mV to reduce crossover distortion – and ‘Robert is Your Father's Brother’ as they say.
👍
Wow valuable Insite indeed would you have any suggestions for upgrades on a Rotel RB-880 Thanks in advance
Hi Woof,
The RB-880 is basically a traditional 3-stage (LTP / VAS / PWR) amp, so the upgrades (which are unfortunately now buried deep into the numerous posts on this thread) will apply.
So you could put in a CM2 current mirror instead of the LTP collector resistors and VAS4 EF-VAS-EF modules for the 2SC1941s.
And then remove the 33k VAS load.
The RB-880 is basically a traditional 3-stage (LTP / VAS / PWR) amp, so the upgrades (which are unfortunately now buried deep into the numerous posts on this thread) will apply.
So you could put in a CM2 current mirror instead of the LTP collector resistors and VAS4 EF-VAS-EF modules for the 2SC1941s.
And then remove the 33k VAS load.
Voltage wise they would perhaps, but isn't the R685 with C619/621 meant to make the supply voltage "cleaner" (than the main rails) for the input pair?
Hi Madis,
You are basically right, but properly balancing the LTP will give a large increase in PSRR that more effectively takes care of rail ripple than filters.
Actually in Rotala #1, I did put in 153R / 100uF (ie. 10Hz LP) rail filters without that getting any better specs than Rotala #2 where I just shorted the 120k.
You are basically right, but properly balancing the LTP will give a large increase in PSRR that more effectively takes care of rail ripple than filters.
Actually in Rotala #1, I did put in 153R / 100uF (ie. 10Hz LP) rail filters without that getting any better specs than Rotala #2 where I just shorted the 120k.
It is all "within reason".
A tightly balanced (matched pair) reduces PSRR because the CMRR is greatly improved. A current source for tail current enhances CMRR as well. Of course, this also greatly reduces distortion too. No matter what your filters are like, and the rest of the circuit, everything hinges on the LTP being matched and in balance for the best performance.
If you effectively regulate the voltage amplifier section, supply ripple has little effect - again within reason.
A tightly balanced (matched pair) reduces PSRR because the CMRR is greatly improved. A current source for tail current enhances CMRR as well. Of course, this also greatly reduces distortion too. No matter what your filters are like, and the rest of the circuit, everything hinges on the LTP being matched and in balance for the best performance.
If you effectively regulate the voltage amplifier section, supply ripple has little effect - again within reason.
Rotel RA-820BX3 ‘Rotala’ (contd.)
I was actually planning not to comment further on the 'filter or no filter' issue, but at least one obvious question remains: Why use a high Johnson noise 120k resistor and 100nF? If you really want, you could get the contentious 13 Hz filter by using, say, 1k2 and 10uF (and then maybe filter both rails).
Or even the 10Hz 154R / 100uF that I tried - or...... whatever floats your boat.🚣♀️
Sort on that topic, I managed to find a copy of the original Otala 1973 paper. Here is the schematic:
You will note that he used a 6k8 and a 10V zener to reduce the c-e voltage over the LTP, as the BCY87 is a max 40V device.
Rotel apparently directly copied the Otala design - inverted it, then added some inexplicably silly things and didn't fix others.
All of which my 14 resistor upgrade in post #1,121 above sorts out.
And in the end you will have a quite pleasantly sounding 'classical' Rotala amp.
I was actually planning not to comment further on the 'filter or no filter' issue, but at least one obvious question remains: Why use a high Johnson noise 120k resistor and 100nF? If you really want, you could get the contentious 13 Hz filter by using, say, 1k2 and 10uF (and then maybe filter both rails).
Or even the 10Hz 154R / 100uF that I tried - or...... whatever floats your boat.🚣♀️
Sort on that topic, I managed to find a copy of the original Otala 1973 paper. Here is the schematic:
You will note that he used a 6k8 and a 10V zener to reduce the c-e voltage over the LTP, as the BCY87 is a max 40V device.
Rotel apparently directly copied the Otala design - inverted it, then added some inexplicably silly things and didn't fix others.
All of which my 14 resistor upgrade in post #1,121 above sorts out.
And in the end you will have a quite pleasantly sounding 'classical' Rotala amp.
Hi AngelP,Rotel RA-820BX3 ‘Rotala’ (contd.)
I got another RA-820 BX3 to see whether the issues I had in ‘Rotala’ #1 were unique for that particular amp or if there should be some systemic faults in these ‘non-Curtis’ designs.
The first thing of notice is that this BX3 ‘Rotala’ #2 has a completely different pcb and even some circuitry differences – without any indicative model name change, not even a “Mk.II”.
View attachment 1376431
Well, that became clearer after a bit of de-dusting:🌪️
View attachment 1376433
After some repairs of faults (yes, I bought it “with some problems” on ebay for £25). I turned it on and it ran just as flaming hot as Rotala #1.
The VAS to about 90 ºC and the 2W 910R driver base resistors got up to a whopping 120 ºC!.🔥
I actually found that these two resistors in the left channel had burned out (!), so the driver transistors only acted as b-e diodes.
The output offsets were -680mV and -595mV (!!) – which could not be trimmed out, but at least they seemed to be stable!
So I decided to first keep the right channel original as reference, and do the required changes to tame the temperature and offsets in the left channel.
The 'bare minimum' upgrade went precisely as smooth as expected. It did take me several hours as I (over-)carefully did a full DC, AC, temp and scope test after each individual change. All temperatures now under 55 ºC and the offset stable at zero with the trimmer in the centre position. No instability whatsoever. Very limited output change to varying temperatures – i.e. everything completely ‘normal’ for a Rotel. THD down at 0.020%, mainly 3rd.
So, if you should be so (un-)fortunate to purchase or be gifted one of the following ‘Rotala’ amps:
RA-820BX3 (version #1 and #2)
RA-820BX4
RA-840BX3
RA-840BX4
RA-940BX
RA-960BX
(there could be more, perhaps someone in this thread can add to this list)?
- you will have to take action because you will have issues of unwanted high output offset and suicidal pcb overheating as these amps all use the exact same Rotel version of the old Otala design.
So, as an absolute minimum you should immediately heat up the old soldering iron and get 14 inexpensive resistors (all 1% MF 1/4W):
4x 270R; 4x 10k; 4x 1k and 2x 82R.
Then replace as follows:
View attachment 1376437
The 270 and 10k’s will remove most of the offset by balancing the LTP (and significantly reduce THD in the process). Trimming to zero is now easy with VR603/4.
The 1k’s will reduce the VAS current and bring the VAS temperature down from over 90 ºC.
The simple 82R (replacing the two big 910Rs) eliminates the heat they produce while doing the same job.
And you could then also remove C613/4 and short out R685 – the LTP’s 2SA1016s can easily take the full rail voltage.
Then trim the bias up to 4mV to reduce crossover distortion – and ‘Robert is Your Father's Brother’ as they say.
Thanks for your great work on this rather unusual amplifier - I am looking forward to starting work on my old RA-820BX4 very soon!
However, I have some questions regarding some of the changes you have made.
Checking the design on simulation with R625, R627, changed to 1k in value from 1K8 (to reduce heat in the VAS stage) seems to effect the overall gain/headroom of the amplifier. Figures I get are 26v peak output with a level of 365mv in - equivalent of around 42 watts into 8 ohms before any losses.
With R625, R627, changed to 1k severe clipping occurs with the same input - reducing it down to 280mv gives 17.6v output, or 19 watts into 8 ohms.
If you could check this on your modified circuit I would be very pleased.
Also, on your "basic minimum" circuit of changes, you would need to change R605 to 2K15 in value for the changes to work (according to simulation) or gross overload in the first stage occurs. Changing resistor R685 in the negative supply back to 120K would stop this and make the circuit work, but this would be a backward step I think. If you could check these things out on your modified amp and put me right, I would be very grateful!
Hi Paul,
That are some very good observations, thank you.
I must admit that I never cared about the wattage of these amps (my 8W Hiraga into my effective Angelhorns will readily blow the pictures off the walls), I mainly concentrated on stopping the Rotalas' destructive overheating and offset issues while reducing the THD.
And, upon re-opening the amps - yes, I had already changed the R605/6 to 2k15 on both. So, the 14 resistor 'bare minimum' upgrade should actually be 16 resistors.
Changing from 8k2 to 2k2 does increase the LTP current to 560uA - still a modest 280uA through each.
Incidentally, compared to the original Lohstrom and Otala design, Rotel did reduce the DC currents in all stages, except the VAS:
I will look into the gain/headroom and wattage issue as soon as I can find the time.
That are some very good observations, thank you.
I must admit that I never cared about the wattage of these amps (my 8W Hiraga into my effective Angelhorns will readily blow the pictures off the walls), I mainly concentrated on stopping the Rotalas' destructive overheating and offset issues while reducing the THD.
And, upon re-opening the amps - yes, I had already changed the R605/6 to 2k15 on both. So, the 14 resistor 'bare minimum' upgrade should actually be 16 resistors.
Changing from 8k2 to 2k2 does increase the LTP current to 560uA - still a modest 280uA through each.Incidentally, compared to the original Lohstrom and Otala design, Rotel did reduce the DC currents in all stages, except the VAS:
| Original | Rotala | AP upgrade | |
| 1st stage: | 200uA | 150uA | 560uA |
| 2nd stage | 4mA | 2mA | 2mA |
| 3rd stage | 20mA | 23mA | 6mA |
| 4th stage | 300mA | 11mA | 20mA |
I will look into the gain/headroom and wattage issue as soon as I can find the time.
Last edited:
Ok, did a quick test on the Rotala with my favourite sim tools: Solder and scope.
Left the 1k resistors in place and kept the current at 2mA - clipping starts at 23Vp, ie. 16Vrms which gives 16x16 / 8 = 32W
Increased the current to 4mA by changing R619 to 301R - clipping still starts at 23Vp
Lifted one pin and added 820R to the 1k to go back to 1k8 - clipping still starts at 23Vp
This seems to confirm that the changes in 2nd stage collector resistors do not change any gain/headroom of the amp, as should not be expected.
Note that I am working on the RA-820BX3 which is a 30W amp. I believe the BX4 is a 40W amp?
I am not sure what could have gone wrong in your simulation - was the 17.6V readout actually the rms value? That would give you 39W?
Left the 1k resistors in place and kept the current at 2mA - clipping starts at 23Vp, ie. 16Vrms which gives 16x16 / 8 = 32W
Increased the current to 4mA by changing R619 to 301R - clipping still starts at 23Vp
Lifted one pin and added 820R to the 1k to go back to 1k8 - clipping still starts at 23Vp
This seems to confirm that the changes in 2nd stage collector resistors do not change any gain/headroom of the amp, as should not be expected.
Note that I am working on the RA-820BX3 which is a 30W amp. I believe the BX4 is a 40W amp?
I am not sure what could have gone wrong in your simulation - was the 17.6V readout actually the rms value? That would give you 39W?
Hi AngelP,
Thanks for checking this on your amplifier.
The RA-820BX4 is the same rated power as the 820-BX3 and the power rails measure about +- 34v off load.
The results I measure on MicroCap 10.2 simulation for 1k resistors are 18v peak output, which is 12.7v RMS which gives 12.7 x 12.7 /8 = 20 watts.
I will double check my circuit configuration for errors. Just out of interest what was your input level for 32W output on your amp?
Thanks for checking this on your amplifier.
The RA-820BX4 is the same rated power as the 820-BX3 and the power rails measure about +- 34v off load.
The results I measure on MicroCap 10.2 simulation for 1k resistors are 18v peak output, which is 12.7v RMS which gives 12.7 x 12.7 /8 = 20 watts.
I will double check my circuit configuration for errors. Just out of interest what was your input level for 32W output on your amp?
Just measured gain of all channels in rb-976MK2. For most channels imbalance is about 0.1-0.2dB but for one channel difference if closer to 1dB. Who is to blame for it? Feedback resistors?
Hi AngelP,
The gain of your amp seems to be different compared to a review by Paul Miller in August 1990! He managed to get 39.4 watts into 8R with a 272mv in to the CD/Aux input. Rotel are usually very conservative with their specs so 32 watts seems a little low for one channel driven into 8 ohms?
Did you replace R601, R603, back to their original positions from your first amp, as this would change the gain in the shunt feedback configuration, unless you re-scaled R653, R657, accordingly?
Attachments
Hi Bacek,
Yes.
Possibly the feedback to common capacitor as well.
paulmc49,
Manufacturers normally specify power into 8R below a certain distortion. They are only slightly conservative and make the rating so that most of their units will exceed it. Line (mains) voltage has more to do with output power than anything else. In other words, the power supply rail voltage determines output power as long as protection isn't acting. At 8R it is not active.
As for gain, running production changes are not uncommon. Many units use 5% resistors for feedback as well, so there will be some variance.
Yes.
Possibly the feedback to common capacitor as well.
paulmc49,
Manufacturers normally specify power into 8R below a certain distortion. They are only slightly conservative and make the rating so that most of their units will exceed it. Line (mains) voltage has more to do with output power than anything else. In other words, the power supply rail voltage determines output power as long as protection isn't acting. At 8R it is not active.
As for gain, running production changes are not uncommon. Many units use 5% resistors for feedback as well, so there will be some variance.