Please use the "Cabirio-BJTmodels.txt" from post #1048 above. If you got it from my other thread, they aren't included in that one.
Correct you are, now the sim passes ok.
You should use versioning or date in your models file name 🙂
Rotel RA-820BX3 VAS Stage 3 Changes:
Then on to Stage 3 and the finger roasting 90ºC Q617/619 VAS. There is no room for suitable heatsinks, so the only remedy is to reduce the standing VAS current.
I know, first violation of the revered Otala design, but necessity always trumps veneration.
And the Rotel engineers seemingly also recognised this problem, as they chose to reduce the current to 13mA, down from Otala’s original design of 20mA.
(Educational part: The VAS current is defined by the second LTP, more specifically the current through the 1k8 resistors R625/627. This determines the base and emitter voltages of Q617/619 and hence the current through R673/675, ie. the VAS current – if you get the calculations right 😊)
Reducing the 1k8s to 1k gave a 6mA VAS current with a large heat decrease.
I replaced the well cooked 2SD600’s with a matched pair of KSC3503 and the current mirror Q613/615 with a matched pair of KSA1381- as I happened to have a tube of each in stock.
The VAS temperature dropped down below 50ºC, much healthier! And it looks a bit more tidy, too.
I probably should have taken a RTA measurement at this point to check that the changes so far have had very limited effect on the THD, but I got a bit carried away and had already started on the first LTP stage and its imbalances.
Then on to Stage 3 and the finger roasting 90ºC Q617/619 VAS. There is no room for suitable heatsinks, so the only remedy is to reduce the standing VAS current.
I know, first violation of the revered Otala design, but necessity always trumps veneration.
And the Rotel engineers seemingly also recognised this problem, as they chose to reduce the current to 13mA, down from Otala’s original design of 20mA.
(Educational part: The VAS current is defined by the second LTP, more specifically the current through the 1k8 resistors R625/627. This determines the base and emitter voltages of Q617/619 and hence the current through R673/675, ie. the VAS current – if you get the calculations right 😊)
Reducing the 1k8s to 1k gave a 6mA VAS current with a large heat decrease.
I replaced the well cooked 2SD600’s with a matched pair of KSC3503 and the current mirror Q613/615 with a matched pair of KSA1381- as I happened to have a tube of each in stock.
The VAS temperature dropped down below 50ºC, much healthier! And it looks a bit more tidy, too.
I probably should have taken a RTA measurement at this point to check that the changes so far have had very limited effect on the THD, but I got a bit carried away and had already started on the first LTP stage and its imbalances.
Excellent!
Yeah, something that works is far more valuable than something that doesn't and is burned past reliable repair any day!
I'm going to bet there is a "happy spot" for Vas current depending on the design. Running things hotter does not improve anything in my experience.
Yeah, something that works is far more valuable than something that doesn't and is burned past reliable repair any day!
I'm going to bet there is a "happy spot" for Vas current depending on the design. Running things hotter does not improve anything in my experience.
Rotel RA-820BX3 LTP Stage 1 Changes
There is no really reason why the R607/609 LTP emitter resistors should differ by this much, so I changed R607 to 470R (to match R609) , and replaced the VR613 470R trimmer with a 100R – at least for the time being. So, I could still zero the offset, although it drifted slowly up and down as before.
Having already replaced the LTP with matched KSA992s, I also removed C613 and changed R615 and 671 to 10k and 12k to straighten up the lopsided Stage 1 to a better balanced stage.
I decided to call this ‘Upgrade 1’ and ran an RTA:
Upgrade 1 trace (green) over the original (black).
Note that the odd harmonics have all dropped significantly, the previous dominant 3rd by a whopping 12dB and the 5th even more!
Overall THD was now down to 0.038% - already ⅓ of what it was - without any major original circuit design changes!
Seems like things are heading in the right direction. 👍
There is no really reason why the R607/609 LTP emitter resistors should differ by this much, so I changed R607 to 470R (to match R609) , and replaced the VR613 470R trimmer with a 100R – at least for the time being. So, I could still zero the offset, although it drifted slowly up and down as before.
Having already replaced the LTP with matched KSA992s, I also removed C613 and changed R615 and 671 to 10k and 12k to straighten up the lopsided Stage 1 to a better balanced stage.
I decided to call this ‘Upgrade 1’ and ran an RTA:
Upgrade 1 trace (green) over the original (black).
Note that the odd harmonics have all dropped significantly, the previous dominant 3rd by a whopping 12dB and the 5th even more!
Overall THD was now down to 0.038% - already ⅓ of what it was - without any major original circuit design changes!
Seems like things are heading in the right direction. 👍
Looking at the graph the 100 Hz noise seems to start to dominate should you succeed in reducing the harmonics any further.
That's just mains pickup plus rectification noise.
Improving voltage regulators and supply isolation might help some. That fish can be fried once the amp is sorted.
It's hard to read the graph to see how bad it is (or isn't). I use -0.9 dBFS as my "0dB" ref level (2.83 V), which is 1 watt into 8R0. That way you can compare easily.
Improving voltage regulators and supply isolation might help some. That fish can be fried once the amp is sorted.
It's hard to read the graph to see how bad it is (or isn't). I use -0.9 dBFS as my "0dB" ref level (2.83 V), which is 1 watt into 8R0. That way you can compare easily.
No, the 50/100Hz harmonics fall away rather nicely, almost none above 500Hz. And as I have posted before, my test setup seems to be overly sensitive to these mains related frequencies.
Yes, I should have done something about it, but with the availability of (almost) affordable devices like the QuantAsylum QA403 - I don't know
Here, I am mainly using these RTAs as comparison means to check whether changes makes things better or worse. Without forking out £££s.
You may start hearing that 100 Hz hum if you make the rest of the amp too good 🙂
You should try RMAA - it may give you a better overview of the device state than just RTA.
Also easy to compare the before and after states numerically:
You may start hearing that 100 Hz hum if you make the rest of the amp too good 🙂
IF the measurement is true. One should always apply a healthy amount of scrutiny to what instruments tell you, in particular if it is not what you expected.
I have upgraded over hundred Rotels and I can say for certain that there is no audible hum. Nor can any hum be seen on oscilloscope traces.
What you see could be a hum loop somewhere in my RTA setup or more probably electromagnetic radiation from the power supply into the RTA probes or input. An INA stage before the Xonar7 could possibly clean up the RTA trace - and I will get round to that - at some point. 😉
And to follow changes, I definitely prefer a graphical display over a list of numbers.
The REW software does provide all these calculated values, but IMHO the trace provides an instant and more complete view of what is going on.
IF the measurement is true. One should always apply a healthy amount of scrutiny to what instruments tell you, in particular if it is not what you expected.
I have upgraded over hundred Rotels and I can say for certain that there is no audible hum. Nor can any hum be seen on oscilloscope traces.
What you see could be a hum loop somewhere in my RTA setup or more probably electromagnetic radiation from the power supply into the RTA probes or input. An INA stage before the Xonar7 could possibly clean up the RTA trace - and I will get round to that - at some point. 😉
And to follow changes, I definitely prefer a graphical display over a list of numbers.
The REW software does provide all these calculated values, but IMHO the trace provides an instant and more complete view of what is going on.
RMAA provides traces too 🙂
Numbers give you a quick summary of changes in different aspects.
I use multi instrument, but for most things I run ARTA (also paid). Try the free version Per.
I find that at the levels we measure, hum is all too common. It seems to be related to the computer connection also. I did buy an audio isolation transformer from Amazon. I tested a few and this came back with the lower distortion. WinSinDA RCA-2
https://www.amazon.ca/Isolator-Alum...efix=rca+isolation+transformer,aps,77&sr=8-31
This has worked well to get a truer picture of supply components in equipment.
I find that at the levels we measure, hum is all too common. It seems to be related to the computer connection also. I did buy an audio isolation transformer from Amazon. I tested a few and this came back with the lower distortion. WinSinDA RCA-2
https://www.amazon.ca/Isolator-Alum...efix=rca+isolation+transformer,aps,77&sr=8-31
This has worked well to get a truer picture of supply components in equipment.
This is drifting OT but in REW you can apply the calibration of your measurement loopback chain and in RMAA you can deduct one measurement (your gear loopback) from the current results.
Both methods should eliminate the environmental impact to measurement result.
Agreed, drifting a little. But I'd like to address the concerns about hum and maybe help with future measurements for all members.
Eliminating the test environment hum at the source is by far the easiest thing to do. Many test programs either don't allow you to remove residuals, or it is an inconvientant next step. I'd rather just plug something in that greatly reduces the problem if it arises.
I'm looking at a USB common mode filter as well that another member designed. Parts on order.
Returning to regularly scheduled programming.
-Chris
Eliminating the test environment hum at the source is by far the easiest thing to do. Many test programs either don't allow you to remove residuals, or it is an inconvientant next step. I'd rather just plug something in that greatly reduces the problem if it arises.
I'm looking at a USB common mode filter as well that another member designed. Parts on order.
Returning to regularly scheduled programming.
-Chris
Rotel RA-820BX3 Offset Stability Problem
OK, we have managed to improve this Rotel, (which is really not a Rotel, but a a copy of a 1973 Otala design).
I guess that we could call it an ROTALA?😆
Anyway, the large output offset instability still puzzled and concerned me. It is something I never have had to sort out before.
To recap, this amp output was drifting back and from + and -700mV (!!) - so much that when I added a speaker protection circuit there were an incessant on/off clicking so I had to disable it.
I already had an indication that it mainly came from somewhere in the first LTP and I had already checked that the current sources Q601 and Q607 were ok and stable.
The original Otala design used 200μA and 4mA respectively through the first two LTP stages – seemingly Rotel decided to change that to 164μA and 2mA.
So I assume that there is some precedence and wiggle-room to work within without dramatic original design violations.
Using a BJT LTP collector current of just 82μA in each is not a problem per se, but it does mean that the base current Ib = Ic / (1+hFE) is well down in the nanoamp range. (Example: 82μA / (1+300) = 272nA)
This is down where you could start to worry about leakage currents on the pcb, in particular the phenolic FR-2 and not least any flux residues or dust left on it. (Of which there can be plenty in an old Rotel).
Transistor specs also start to change with very low Ic so you have both internal as well as external things to consider.
In my experience you definitely need to control the environment of low current devices, and an open audio amp board is clearly not the best place for this.
So, what to do about the problem?
I decided to first try to increase the Stage1 current, experimentally by hooking a resistance decade box across R605 and paralleling it with successfully lower values to increase the current source output - while testing each setting with a spray of freeze on the LTP and watching the amp offset response.
And indeed, at about 2k7 in parallel with the 8k2 giving ~600μA in the current source (~300μA in each LTP transistor) - the offset became immune to a quick freeze douche of the first LTP – ok, almost. (Note that I had already put in a closely matched pair of KSA992’s).
The best result was reached by replacing the 8k2 R605 with a 2k15, increasing the stage current to just over 550μA.
(That may well have to be changed again as we move on, but at least it is a good starting point.)
The output offset is now much more stable, and less depending on temperature - the output mostly stays within +/-100mV.
Not bad at all...... but not perfect by any measure.
Stay tuned.
OK, we have managed to improve this Rotel, (which is really not a Rotel, but a a copy of a 1973 Otala design).
I guess that we could call it an ROTALA?😆
Anyway, the large output offset instability still puzzled and concerned me. It is something I never have had to sort out before.
To recap, this amp output was drifting back and from + and -700mV (!!) - so much that when I added a speaker protection circuit there were an incessant on/off clicking so I had to disable it.
I already had an indication that it mainly came from somewhere in the first LTP and I had already checked that the current sources Q601 and Q607 were ok and stable.
The original Otala design used 200μA and 4mA respectively through the first two LTP stages – seemingly Rotel decided to change that to 164μA and 2mA.
So I assume that there is some precedence and wiggle-room to work within without dramatic original design violations.
Using a BJT LTP collector current of just 82μA in each is not a problem per se, but it does mean that the base current Ib = Ic / (1+hFE) is well down in the nanoamp range. (Example: 82μA / (1+300) = 272nA)
This is down where you could start to worry about leakage currents on the pcb, in particular the phenolic FR-2 and not least any flux residues or dust left on it. (Of which there can be plenty in an old Rotel).
Transistor specs also start to change with very low Ic so you have both internal as well as external things to consider.
In my experience you definitely need to control the environment of low current devices, and an open audio amp board is clearly not the best place for this.
So, what to do about the problem?
I decided to first try to increase the Stage1 current, experimentally by hooking a resistance decade box across R605 and paralleling it with successfully lower values to increase the current source output - while testing each setting with a spray of freeze on the LTP and watching the amp offset response.
And indeed, at about 2k7 in parallel with the 8k2 giving ~600μA in the current source (~300μA in each LTP transistor) - the offset became immune to a quick freeze douche of the first LTP – ok, almost. (Note that I had already put in a closely matched pair of KSA992’s).
The best result was reached by replacing the 8k2 R605 with a 2k15, increasing the stage current to just over 550μA.
(That may well have to be changed again as we move on, but at least it is a good starting point.)
The output offset is now much more stable, and less depending on temperature - the output mostly stays within +/-100mV.
Not bad at all...... but not perfect by any measure.
Stay tuned.
Rotel RA-820BX3 Offset Stability Problem (contd.)
Thus emboldened by the success of increasing Stage1 collector current, I moved on to try the same in the second LTP Q609/611.
Here, you have to be a bit careful, as the current through this stage also determines the following VAS stage current (see post #1063). So if you overdo it, there will be VAS smoke on the water and maybe even fire in the sky.
I first increased the Stage2 to 3mA and tried the freeze spray test. No obvious improvement. 🤔
I could of course (again) reduce the emitter resistor values and increase the current even further without roasting the VAS, but then I remembered that I hadn’t checked out the old 2SA1016 pair.
As mentioned above, the hFE of the first LTP was mismatched by 13% - the second by a whopping 41% !!
So in with a new matched pair of KSA992s and things improved immediately. The temp drifting was reduced further, but there was a constant offset of -165mV.
First, I centered the 100R offset trimmer in Stage1 (I already had matched the two 470R emitter resistors R607/609) and then I tried changing the value of one of Stage2’s 470R resistors by paralleling it with kOhm resistors. Thus changing from 470R to the equivalent of 459R (11 ohms!) I could zero the constant offset. Now, I didn’t have that value in stock for R623, so I just chose 464R and soldered a 47k in parallel with it.
Then I thought – why do I need the offset trimmer now when it is centered? So I removed it and shorted the current source to the emitter resistors accordingly. The first input LTP is now as balanced and linear as I can get it, the second almost so - what happens when I power the amp up from the internal supply?
(I probably should mention that I always power the amp from two 30V lab supplies with the current limiting set conservatively when I play around with the circuitry this way. It is not strictly necessary, but it does reduce the crying and cursing a lot).
With high hopes for a better performance, I hooked the RTA measurement up and ......oops, the 1kHz test tone down many dB and a heavy amount of 50Hz harmonics – what??😵
I quickly tried to hook up a music player and headphones…. Only a whisper of music through the channel with the volume set at max!
Was this just ‘A Series of Unfortunate Events’? Or had I freeze sprayed the LTPs so much that they had died from thermal cycling? Or killed the amp in another most exciting way?
Or am I actually an Absolute Plonker? (The most probable explanation).
Stay tuned...
Thus emboldened by the success of increasing Stage1 collector current, I moved on to try the same in the second LTP Q609/611.
Here, you have to be a bit careful, as the current through this stage also determines the following VAS stage current (see post #1063). So if you overdo it, there will be VAS smoke on the water and maybe even fire in the sky.
I first increased the Stage2 to 3mA and tried the freeze spray test. No obvious improvement. 🤔
I could of course (again) reduce the emitter resistor values and increase the current even further without roasting the VAS, but then I remembered that I hadn’t checked out the old 2SA1016 pair.
As mentioned above, the hFE of the first LTP was mismatched by 13% - the second by a whopping 41% !!
So in with a new matched pair of KSA992s and things improved immediately. The temp drifting was reduced further, but there was a constant offset of -165mV.
First, I centered the 100R offset trimmer in Stage1 (I already had matched the two 470R emitter resistors R607/609) and then I tried changing the value of one of Stage2’s 470R resistors by paralleling it with kOhm resistors. Thus changing from 470R to the equivalent of 459R (11 ohms!) I could zero the constant offset. Now, I didn’t have that value in stock for R623, so I just chose 464R and soldered a 47k in parallel with it.
Then I thought – why do I need the offset trimmer now when it is centered? So I removed it and shorted the current source to the emitter resistors accordingly. The first input LTP is now as balanced and linear as I can get it, the second almost so - what happens when I power the amp up from the internal supply?
(I probably should mention that I always power the amp from two 30V lab supplies with the current limiting set conservatively when I play around with the circuitry this way. It is not strictly necessary, but it does reduce the crying and cursing a lot).
With high hopes for a better performance, I hooked the RTA measurement up and ......oops, the 1kHz test tone down many dB and a heavy amount of 50Hz harmonics – what??😵
I quickly tried to hook up a music player and headphones…. Only a whisper of music through the channel with the volume set at max!
Was this just ‘A Series of Unfortunate Events’? Or had I freeze sprayed the LTPs so much that they had died from thermal cycling? Or killed the amp in another most exciting way?
Or am I actually an Absolute Plonker? (The most probable explanation).
Stay tuned...
Last edited:
Hi Per,
Wow! Huge imbalance.
When I see diff pairs, I always match them. The second shouldn't have nearly as big an effect as the primary diff pair, but it does matter.
Your loss of signal will be something simple <hand smacked in to forehead time!>.
Wow! Huge imbalance.
When I see diff pairs, I always match them. The second shouldn't have nearly as big an effect as the primary diff pair, but it does matter.
Your loss of signal will be something simple <hand smacked in to forehead time!>.