I normally don't like bypassing caps at all.
There was some higher frequency noise distorting the signal, like those exaggerated "Zzzz..." from female voices, etc.
My output caps consist of 22uF || 2 x 1.1uF film. I added a 2n2 and those "Zzzz..." sound got reduced and the clarity of sound improved.
Same was found with the caps in front of the regulator.
So I thought that I would apply to the Vref caps. They were 56uF || 1.1uF initially. I added a 2n2 this morning.
Since I changed both the pre-reg and added the 2n2 at the same time, I don't know if that 2n2 works or not.
Why not parallelling the Vref cap? Any technical explanation for it?
My reasoning is that, just as Ikoflexer uses RF devices, if the capacitors do not work at higher frequencies, the goal can not be achieved. Smaller caps work at higher frequencies. I guess the art is to find the right values.
Refer to the Vishay bluebox graph below, the 1.1uF does not work very well at higher frequencies.
An externally hosted image should be here but it was not working when we last tested it.
By the way, how much capacitance do you guys use for the raw supply?
I have the 600V 30A rectifier diodes mounted on aluminium so I could technically use as large capacitance as needed without worrying inrush current damaging any devices. The LR I use are rather small, merely 470uH and 0.5R. The scope shows just under 0.1V ripples (peak to peak) before the regulator.
I have the 600V 30A rectifier diodes mounted on aluminium so I could technically use as large capacitance as needed without worrying inrush current damaging any devices. The LR I use are rather small, merely 470uH and 0.5R. The scope shows just under 0.1V ripples (peak to peak) before the regulator.
Stormsonic,
You are absolutely right. I have tested them and found that multiple bypass really ruined the sound.
More tips? 🙂
Regards,
Bill
Find low leakage current capacitors. Film capacitors are OK, solid polymer (OSCON) is NO NO, lytic some are good, some are not 🙂
Stormsonic,
What low leakage capacitor do you use? my electrolytic is about 3uA.
Do you mind sharing what you have in front of the regulator, i.e. the raw supply?
Regards,
Bill
What low leakage capacitor do you use? my electrolytic is about 3uA.
Do you mind sharing what you have in front of the regulator, i.e. the raw supply?
Regards,
Bill
If we reform our electrolytics at their rated voltage, then does anyone know which types maintain their low leakage when operated long term @ <<rated voltage
I am using Rubycon ZL, ZLG and ZA mixed with Panasonic FC. All of these have the same spec with 3uA+ leakage current. I searched and found when the leakage current is 1uA the manufactures classify them as low leakage, but then those are with ESR/ESL much higher, typically at 1R, versus 0.01R to 0.1R for the low ESR/ESL type. Yes apparently at the Vref location any leakage current introduces noise / distortions.
I never thought that the RAW supply has such a large impact on the sound.
I have tried many various configurations and have not found the optimal. I have had LRCLRCLRCLRC for some time now, and found that any R in series introduces distortions. Without some R (even as low as 0.25R), the LC would create some high frequency noise that can not be measured with my scope but can clearly be heard. So that leaves me the only choice of C only. In that case, the overall presentation of the sound is excellent, but high frequency filtering suffers, and I can hear the edgeness in female voices.
Tried putting a LM317/337 pre-regulator in but it solves some problems but introduces others.
Next I will try putting the snubbing caps across the rectifier diodes again and may increase the C to 15,000uF and see if it works.
Would you use C up to 20,000uF?
I have tried many various configurations and have not found the optimal. I have had LRCLRCLRCLRC for some time now, and found that any R in series introduces distortions. Without some R (even as low as 0.25R), the LC would create some high frequency noise that can not be measured with my scope but can clearly be heard. So that leaves me the only choice of C only. In that case, the overall presentation of the sound is excellent, but high frequency filtering suffers, and I can hear the edgeness in female voices.
Tried putting a LM317/337 pre-regulator in but it solves some problems but introduces others.
Next I will try putting the snubbing caps across the rectifier diodes again and may increase the C to 15,000uF and see if it works.
Would you use C up to 20,000uF?
L might introduce some resonance. A good rule of thumbs is to set a low C before L and then a 3 or 4xC after. I.E. 2,200 (or 3,300) uF + L + 10,000uF
I read about you experiments & measurements of leakage current at diyhifi.
I measured them with procedure similar to THIS, pic on page 7, with 3V batteries. Mostly FC, FM, Elna's and some cheap types like Suntan, Yageo,....
But never at rated voltage and never observed their behavior & leakage over long period of time. I am curious, please let us know which types maintain their low leakage on long term 🙂
Hi,
thanks for the link.
My method is similar but the lowest resolution DMM is 100uVdc.
To get readings across the leakage indication resistor, I use a 10k or 100k resistor.
I have used a regulated supply for the DC source and I think charging discharging currents as the regulated voltage changes with temperature and mains voltage tend to swamp the voltage readings for leakage.
I have never tried using a low voltage battery supply into a high voltage capacitor.
thanks for the link.
My method is similar but the lowest resolution DMM is 100uVdc.
To get readings across the leakage indication resistor, I use a 10k or 100k resistor.
I have used a regulated supply for the DC source and I think charging discharging currents as the regulated voltage changes with temperature and mains voltage tend to swamp the voltage readings for leakage.
I have never tried using a low voltage battery supply into a high voltage capacitor.
Things are looking good.
Last night I (hopefully) traced down where that edgeness of female voices come from. I upgraded my garden variety diode bridge to the Fairchild Steath fast / soft recovery diodes a little while ago. According to the data sheet, no snubbing capacitors are required. From my limited scope, the output from the Steath was obvious much cleaner, so I did not use any snubbing caps.
I added the snubbing caps last night and the wave forms appear to be better again. Subjectively, the edgeness reduced and the sound stage became deeper. I only put in 1.5nF, in comparison to normally recommended 10nF to 100nF for the normal diodes. I need to do some experiments today to find the optimal value.
Last night I (hopefully) traced down where that edgeness of female voices come from. I upgraded my garden variety diode bridge to the Fairchild Steath fast / soft recovery diodes a little while ago. According to the data sheet, no snubbing capacitors are required. From my limited scope, the output from the Steath was obvious much cleaner, so I did not use any snubbing caps.
I added the snubbing caps last night and the wave forms appear to be better again. Subjectively, the edgeness reduced and the sound stage became deeper. I only put in 1.5nF, in comparison to normally recommended 10nF to 100nF for the normal diodes. I need to do some experiments today to find the optimal value.
Now a couple of questions regarding the regulator that need expert advice.
1. What is the minimum load current required for the regulator to regulate?
For a few times I found the regulators did not really "regulate well". I measured up to 0.5V between the load, i.e. opamp supply pin (happened to negative rail only) and the MOSFET output. It was inconsistent - sometimes the problem was there and sometimes not. It really puzzled me. I guess now I have found out the cause. When I took out the opamp from the IC socket, i.e. no load, then the MOSFET output measured correctly at 15V but at the IC socket point it was 14.5V. This suggests to me, that when the opamp current draw is too low (typical some uA from opamps), the regulator won't work.
2. It is a good idea to have a protection diode across the output and input?
What happened if the input of the regulator is shorted while the output capacitors are charged? The voltage at the output is higher than the input. Would it cause any problem to the active devices? If so, perhaps a protection diode, like the one illustrated at the LM317 datesheet, may be helpful, just in case.
1. What is the minimum load current required for the regulator to regulate?
For a few times I found the regulators did not really "regulate well". I measured up to 0.5V between the load, i.e. opamp supply pin (happened to negative rail only) and the MOSFET output. It was inconsistent - sometimes the problem was there and sometimes not. It really puzzled me. I guess now I have found out the cause. When I took out the opamp from the IC socket, i.e. no load, then the MOSFET output measured correctly at 15V but at the IC socket point it was 14.5V. This suggests to me, that when the opamp current draw is too low (typical some uA from opamps), the regulator won't work.
2. It is a good idea to have a protection diode across the output and input?
What happened if the input of the regulator is shorted while the output capacitors are charged? The voltage at the output is higher than the input. Would it cause any problem to the active devices? If so, perhaps a protection diode, like the one illustrated at the LM317 datesheet, may be helpful, just in case.
the CCS limits the current to a fixed value.
You cannot get more than that fixed value.
Some flows in the shunt control circuits.
The remainder is shared between the load and the shunt.
If the load takes none then the shunt will try to take all of that remainder. Power dissipation permitting, either short term or long term. the shunt will regulate with zero load current.
If the output is shorted, the the shunt takes non of that remainder. The load takes it all.
The CCS now drops extra voltage, at the same fixed current, and so runs hotter.
You cannot get more than that fixed value.
Some flows in the shunt control circuits.
The remainder is shared between the load and the shunt.
If the load takes none then the shunt will try to take all of that remainder. Power dissipation permitting, either short term or long term. the shunt will regulate with zero load current.
If the output is shorted, the the shunt takes non of that remainder. The load takes it all.
The CCS now drops extra voltage, at the same fixed current, and so runs hotter.
- Status
- Not open for further replies.