Incoming Voltage Limit
Can this super reg circuit be used with an incoming voltage of 35 volts
assuming the cap voltages are increased ?
Can this super reg circuit be used with an incoming voltage of 35 volts
assuming the cap voltages are increased ?
The error amp is run off the output rail, so it depends on the voltage thereupon.
Try it with the OPA604 which will run on +/- 24 V rails and report back. Tell us if there's smoke.
Here is another question.
Why remote sensing can cause oscillation?
In my implementation, I found that there is no oscillation if the sensing point is wired to the regulator output directly. However, I have no luck with remote sensing. It always oscillates with about 6 cm length of wires, despite only mildly.
The difference is only some small inductance in the sensing wire and the ground sensing wire. The recommended fix is to apply an RC (where R=10, C=10nF) to the sensing wire by shunting to the rail with the 10nF.
But is it because we don't want RF to be input into the error amp? There shouldn't be RF anyway. In that case, there should not be oscillation but resonance only. The fact is that I found mild oscillation. This is hard to understand because Spice simulation does not indicate anything like phase reversal regardless of having the RC or not.
If the issue is the extra inductance reacting to the 120uF cap then a small R in the sensing wire should damp the resonance and there is no need to have an RC.
I have this question because Spice simulation shows that having the RC degrades the output impedance by 20dB between 100kHz and 1MHz. In real life, adding the RC (R=10, C=10nF) completely got rid of the oscillation but the sound was degraded (comparing to not using remote sensing). Reducing the C to 1nF worked with no oscillation and the sound was improved, even better than having no remote sensing, but I could still hear some emphasis at the treble, indicating some mild resonance. Reducing R from 10R to 1R did not work, and the scope showed mild oscillation. I want to experiment to find out the lowest possible values of RC without causing oscillation and resonance. But first, I want to understand what triggers the oscillation.
Why remote sensing can cause oscillation?
In my implementation, I found that there is no oscillation if the sensing point is wired to the regulator output directly. However, I have no luck with remote sensing. It always oscillates with about 6 cm length of wires, despite only mildly.
The difference is only some small inductance in the sensing wire and the ground sensing wire. The recommended fix is to apply an RC (where R=10, C=10nF) to the sensing wire by shunting to the rail with the 10nF.
But is it because we don't want RF to be input into the error amp? There shouldn't be RF anyway. In that case, there should not be oscillation but resonance only. The fact is that I found mild oscillation. This is hard to understand because Spice simulation does not indicate anything like phase reversal regardless of having the RC or not.
If the issue is the extra inductance reacting to the 120uF cap then a small R in the sensing wire should damp the resonance and there is no need to have an RC.
I have this question because Spice simulation shows that having the RC degrades the output impedance by 20dB between 100kHz and 1MHz. In real life, adding the RC (R=10, C=10nF) completely got rid of the oscillation but the sound was degraded (comparing to not using remote sensing). Reducing the C to 1nF worked with no oscillation and the sound was improved, even better than having no remote sensing, but I could still hear some emphasis at the treble, indicating some mild resonance. Reducing R from 10R to 1R did not work, and the scope showed mild oscillation. I want to experiment to find out the lowest possible values of RC without causing oscillation and resonance. But first, I want to understand what triggers the oscillation.
Not using the remote sensing also crossed my mind! What the most important is the sound quality so I will try first whit no sensing and compare to sensing! 😱
It can oscillate because the inductance and capacitance added by the remote wiring causes phase shift in the loop while there is still loop gain. The RC filter decreases the gain at high frequency so it cannot oscillate. The upshot is less regulation at those hi freqs as you measured.
I added the remote sense on request and it can improve regulation at the load, but there is no law saying it sounds bad without! Try it out.
Jan
I think Jan's answer is spot on.
I now reversed back to the original recommended 10nF 10R. I found this gives better sound (than 1nF 10R). I initially changed 2 things at the same time - remote sensing RC and opamp RC compensation and found they degraded the sound but I did not know which did it more. One hour ago I changed the remote sensing RC only and found out the answer. I may still experiment reducing the resistors and see if it improves the sound.
So subjective sound testing matches simulation (almost).
Both simulation and subjective listening showed me that remote sensing makes the sound quality SUBSTANTIALLY better! Without the RC, remote sensing caused oscillation, and to me it happened every single time. With the RC, remote sensing degrades performance only above 100kHz. It does not degrade performance below 100kHz. Without remote sensing, performance degradation is much larger and is on full bandwidth of audio. My simulation results look simular to Jan's measurements published in the original Jung Super regulator Article 3. The difference is huge. I have found that difference in subjective listening.
As for the opamp RC compensation, simulations tell me that it degrades the sound in full bandwidth of the regulator, not just above 100kHz. Subjective listening tells me that the sound became less clean. I still left my much reduced 10pF caps there and the resistors reduced to 10R. The RC doesn't limit the bandwidth any more (practically). But I am dreaming they may still have some effects in providing stability. They definitely don't sound bad.
I now reversed back to the original recommended 10nF 10R. I found this gives better sound (than 1nF 10R). I initially changed 2 things at the same time - remote sensing RC and opamp RC compensation and found they degraded the sound but I did not know which did it more. One hour ago I changed the remote sensing RC only and found out the answer. I may still experiment reducing the resistors and see if it improves the sound.
So subjective sound testing matches simulation (almost).
Both simulation and subjective listening showed me that remote sensing makes the sound quality SUBSTANTIALLY better! Without the RC, remote sensing caused oscillation, and to me it happened every single time. With the RC, remote sensing degrades performance only above 100kHz. It does not degrade performance below 100kHz. Without remote sensing, performance degradation is much larger and is on full bandwidth of audio. My simulation results look simular to Jan's measurements published in the original Jung Super regulator Article 3. The difference is huge. I have found that difference in subjective listening.
As for the opamp RC compensation, simulations tell me that it degrades the sound in full bandwidth of the regulator, not just above 100kHz. Subjective listening tells me that the sound became less clean. I still left my much reduced 10pF caps there and the resistors reduced to 10R. The RC doesn't limit the bandwidth any more (practically). But I am dreaming they may still have some effects in providing stability. They definitely don't sound bad.
You should be able to attach a "quasimodo" to the output of the regulator and "SEE" what ringing there is on the supply rail.
Remote sensing is helpful if you have "eye-squared R" losses -- that's why you see it in the Kepco or Lambda supplies used in the lab.
I am surrendering to Jan and attaching the sense wires to the most power hungry part of the preamplifier instead of the poorest PSRR stage.
I am surrendering to Jan and attaching the sense wires to the most power hungry part of the preamplifier instead of the poorest PSRR stage.
Mark Johnson's standard Quasimodo test jig? How? I am interested. I have a Quasimodo that is ready to use. How to connect it?
To drop incoming voltage under the limits of the AD825, can I put an R/C filter in front of the reg of 100 ohm / 100 uf without affecting performance ?
Why 100 ohms? How did you determine that value? Just curious.
100 ohms / 100uF does not drop DC, except by the load current and that is not defined.
A series zener does that, but as has been pointed out the AD825 is fed from Vout which is not 35V.
Jan
100 ohms / 100uF does not drop DC, except by the load current and that is not defined.
A series zener does that, but as has been pointed out the AD825 is fed from Vout which is not 35V.
Jan
Voltage drop
Jan
My application for the superreg is one of the my ref chip amp designs where I would like to supply the 3 line level op amps ( <50ma - 15 V) with the SR
Available DC is 35V. Friend who is a repair tech threw the 100 ohm resistor value for me to bring it closer to 30 V.
Any suggestions ?
Thx
Jan
My application for the superreg is one of the my ref chip amp designs where I would like to supply the 3 line level op amps ( <50ma - 15 V) with the SR
Available DC is 35V. Friend who is a repair tech threw the 100 ohm resistor value for me to bring it closer to 30 V.
Any suggestions ?
Thx
IF the load is a constant 50mA then 100 ohms would drop 5V. But that load is a max design value and not exactly or always that.
Better is a 4.7V zener in series, that always drops 4.7V independent (pretty much) of the load.
But what is the problem with 35V?
Jan
Better is a 4.7V zener in series, that always drops 4.7V independent (pretty much) of the load.
But what is the problem with 35V?
Jan
Jan
Looks I was worried about nothing since the AD825 is fed from the Vout.
I'll just feed the reg the 35V
Thanks for putting back on the right road !!
Looks I was worried about nothing since the AD825 is fed from the Vout.
I'll just feed the reg the 35V
Thanks for putting back on the right road !!
Heatsink Requirement
If I run 200mA current, will a 30 deg C/W heatsink be sufficient?
But before that question can be answered, we need to establish the voltage drop across the Jung Supereg. Is it 1.5VDC or 2VDC? - assumed that the input is DC, not AC.
For 2VDC voltage drop, dissipation will be 2V x 0.2A = 0.4W. For a 30 deg C/W heatsink, the temperature will rise 0.4 * 30 = 12 deg C above ambient temperature.
That should be very safe, or not?
If I run 200mA current, will a 30 deg C/W heatsink be sufficient?
But before that question can be answered, we need to establish the voltage drop across the Jung Supereg. Is it 1.5VDC or 2VDC? - assumed that the input is DC, not AC.
For 2VDC voltage drop, dissipation will be 2V x 0.2A = 0.4W. For a 30 deg C/W heatsink, the temperature will rise 0.4 * 30 = 12 deg C above ambient temperature.
That should be very safe, or not?
With the Jung Supereg, can we replace the resistors with more common E24 values, e.g. from 249R to 270R, from 4k99 to 5k6R, from 499R to 1k, from 1k to 2k?
When the FET input AD825 is used the slight increase in input impedance makes no impact on the overall noise performance.
When the FET input AD825 is used the slight increase in input impedance makes no impact on the overall noise performance.
Last edited: