Well Marce, I think you will find:
1N4148: very fast, but no soft recovery, will have measurable overshoot. explains the 'brightness', if you filter this out you will lose more than you bargained for?
Vishay hexfred: ultrasoft recovery, made to decrease overall part cost of psu. my guess is that recovery is too soft for audio purposes. my guess is that Fairchild stealth has the same problem.
Cree SiC: [FONT=Verdana,Verdana][FONT=Verdana,Verdana]Zero Reverse Recovery Current, [/FONT][/FONT][FONT=Verdana,Verdana][FONT=Verdana,Verdana]Zero Forward Recovery Voltage, [/FONT][/FONT][FONT=Verdana,Verdana][FONT=Verdana,Verdana]High-Frequency Operation, low capacitance. it is simply a very good diode
[/FONT][/FONT]
1N4148: very fast, but no soft recovery, will have measurable overshoot. explains the 'brightness', if you filter this out you will lose more than you bargained for?
Vishay hexfred: ultrasoft recovery, made to decrease overall part cost of psu. my guess is that recovery is too soft for audio purposes. my guess is that Fairchild stealth has the same problem.
Cree SiC: [FONT=Verdana,Verdana][FONT=Verdana,Verdana]Zero Reverse Recovery Current, [/FONT][/FONT][FONT=Verdana,Verdana][FONT=Verdana,Verdana]Zero Forward Recovery Voltage, [/FONT][/FONT][FONT=Verdana,Verdana][FONT=Verdana,Verdana]High-Frequency Operation, low capacitance. it is simply a very good diode
[/FONT][/FONT]
Myself.... interesting in hearing if the schottky makes a difference, or are they best for small signal PSU's...
+10
Why not answer the question , can you hear a difference..? with anything ..?
Can you hear a difference ? ....
The PSU is the most important part , it will set the stage of what you hear for sound...
Link...?
+10
Maybe .....anyway a difference is detected and the detection once reconized will be audible, so their will be a difference and a preference...
Cree SiC at RS components, you can also find the datasheet there:
Buy cree sic online from RS Components
Buy cree sic online from RS Components
The 4148 turns off in about 1.5 nanoseconds when it's current is driven from .5 amps forward to 1 amp reverse in 250 picoseconds. (low duty cycle testing..) Trr is usually the same over 5 orders of magnitude current change, and indeed, the 4148 also switches off in 1.5 nanoseconds when driven at 5 milliamp forward, then 10 milliamps reverse in 250 pico.
"Clean switch off" is of course, in the eye of the beholder. Circuit geometric design which affect parasitic inductances and capacitances can easily lead one to believe it turns off clean. But with a good test setup, these puppies shut off in less than half a nanosecond..poor layout could easily make life interesting when currents slew at 2 amperes per nanosecond...and a scope probe capacitance can easily trash the higher harmonics generated.
If you can hear the difference with different diodes (!) then your PSU is modulating the output, so there should be observable data, and the PSU design is rubbish, IMO
Actually...two things.
1. It could easily be that the supply is good, but ground layout may not be adequate for EMC problems.
2. If I remember correctly, I read somewhere (this forum I believe), Nelson Pass at some point found that the supply diodes caused an audible change.. This is third or fourth hand info, so perhaps Nelson could clarify??
Cheers, jn
edit: It is not easy to stop 2 amp/nSec current transients. Electrolytics do not exist at those speeds as a result of geometry.
Last edited:
John,
Nice to see you checking things out.
I have found brand of diode makes a huge difference. As mentioned I have some old 1N4004s that created switch off noise around twice the supply voltage! These were so bad they could cause problems even with TTL circuitry!
The secret to cleaning up 1N4148s and almost all diodes is of course a lossy inductor. I am also not opposed to a small capacitor across the diode. I kind of assumed that most high end designs use those techniques. To keep the DC from saturating the inductor the choice is air core or balanced bifilar.
ES
Nice to see you checking things out.
I have found brand of diode makes a huge difference. As mentioned I have some old 1N4004s that created switch off noise around twice the supply voltage! These were so bad they could cause problems even with TTL circuitry!
The secret to cleaning up 1N4148s and almost all diodes is of course a lossy inductor. I am also not opposed to a small capacitor across the diode. I kind of assumed that most high end designs use those techniques. To keep the DC from saturating the inductor the choice is air core or balanced bifilar.
ES
Last edited:
Let me add a few more pennies
There are a few other diodes worth mentioning after my ventures in diode land:
Vishay 1N5062 avalanche diode (the grey glass bead, do an ebay search): softer mids, bit less detailed, but nevertheless a very good 'middle of the road' diode with excellent hi end extention. will be an upgrade in cheap or harsh equipment.
(TSC) 1N5819 Schottky diode: cheap upgrade for cheap equipment, but sounds hyped. may leave a nice first impression, but will fatigue in the long run. if you need some extra high end in a dull system these are woth to try, they cost pennies.
There are a few other diodes worth mentioning after my ventures in diode land:
Vishay 1N5062 avalanche diode (the grey glass bead, do an ebay search): softer mids, bit less detailed, but nevertheless a very good 'middle of the road' diode with excellent hi end extention. will be an upgrade in cheap or harsh equipment.
(TSC) 1N5819 Schottky diode: cheap upgrade for cheap equipment, but sounds hyped. may leave a nice first impression, but will fatigue in the long run. if you need some extra high end in a dull system these are woth to try, they cost pennies.
Nobody expects the Spanish Inquisition!
(I'm still not sure how one "hears" capacitors or resistors or whatever- they generally make no sound unless they're piezoelectric, but all of that is irrelevant anyway when it comes to switching noise)
For anyone interested:
http://www.vincotech.com/fileadmin/user_upload/articles/Bodo'sPowerSystems_May2008_.pdf
A good piece on SiC diodes...including comparisson to FRED. For me this explains a LOT. Now tell me again that I'm crazy...
http://www.vincotech.com/fileadmin/user_upload/articles/Bodo'sPowerSystems_May2008_.pdf
A good piece on SiC diodes...including comparisson to FRED. For me this explains a LOT. Now tell me again that I'm crazy...
Last edited:
Still a bad design then as the layout is critical IMO, more-so when you start playing with SMPS.
We agree. I do find however, that many designers wil make a fantastic black box supply, but put that box into the equipment in such a way as to compromise the EMC performance of the sum.
Quite a few very very large manufacturers will have incredibly good power supply, low level analog, digital, and hf power designers..partition the workload based on expertise, and hope the final product meets goals. When it doesn't, they call in people like Tom Van Doren to troubleshoot.
I also believe they shouldn't. Unfortunately, the equipment rarely asks my beliefs.
Consider an amp with beefy lytics. At 5 Khz, what is the lytic capacitance, ESL, and RSL? Is the layout preventing that audio from proceeding to the line cord via the diode bridge? Or, will the diodes form a PWM based switch, allowing musical content to travel into the line, or conversly, line hash into the lytics.
This is easy enough to test. Monitor the line cord for the music spectra..
Measurement of stuff getting into the equip from external during high power operation when the diodes are on the longest...THAT is not easy.
Cheers, jn
Typical bridge rectification produces haversine current pulses on the line, which are odd harmonics.
If you pushed DC at the output terminal, you'll see prime and odd harmonics on the line draw.
As the output goes from DC to frequency, that draw will be amplitude modulated by the power draw and pulse width modulated by the instantaneous power level being drawn by amp...this by the on time of the diodes.
Eventually at the higher frequencies, the on time of the diodes will be only power line modulated, and that on time will gate high frequencies onto the line, depending on the layout of the supply and the ESR/ESL, and capacitance vs frequency characteristics of the supply capacitors.
All this explanation is fine and dandy, but nothing compares to looking at the draw current waveform with a hf mag clamp on probe.
That is a synopsis of the audio to line connection only. Going the other way, line to audio, has that path in reverse, but in addition, there is the safety ground and/or neutral loop formed with the IC's that is an additional concern.
Tis a big hole in the understandings of IC/PC/layout/ground loop technology when unbalanced systems are used.
And this black box we call a "power amp" has a power gain of roughly....
1kw (10 power 3 watts) output for 1 volt into a 10k resistor (10power -4 watts).
A box with a power gain of 10 power 7, and still people care little about in/out coupling between all the ports....my goodness.
Cheers, jn
On capacitor input filters, yes, but as I understand it inductor input filters ideally have no such problem - although there might be a bit of harmonics due to transformer center-tap imbalances and such.
Agreed. I was considering only a hard full wave setup in my discussion of haversine generated line currents.
The supply geometric layout for inductor input filters can still bite ya for three port interaction via ground loops, but the inductor plays nicer with the line.
jn
Actually, a choke input filter keeps the transformer current quite stable, and HF spikes from capacitor charging are not an issue providing the choke is above the critical value.
LC filters aren't the ideal solution for all cases, as you point out, but for class-A amplifiers it's hard to argue with a nice fat inductor as the first filter element.
How does one determine the size of the inductor needed for clc type supply and is this only benificial for class-a type amps or for all in general..?
Regards,
Power Supply rectifier diodes
I replaced the garden-variety diodes in the power supply rectifier section of a Musical Fidelity XCAN V3 headphone amp with HEXFRED and afterwards measured about 1.5 dB less noise in the output of the amplifier into a 150 ohm load. The amplifier did not sound any different to me.
I did not use A weighting or any other filtering in my measurement (probably should have, in retrospect) and I measured using equipment which was flat to 20 MHz....
I suspect that the decline in noise was equal energy per octave, but as there are more ultrasonic and RF octaves than audio octaves I think that the reduction in noise was mostly at frequencies well beyond my hearing. I should have done a spectral analysis of the residual noise.
Reducing RF spuriae from diode switching in the AC supply could have an impact on audio - many devices used for voltage or current gain such as bipolar transistors and FETs can show nonlinear behavior when fed signals like these, which could have an impact on the audio they are amplifying. COULD - it's a pretty big "IF" here.
And you'd think that none of this would make it past the filter caps, but then again noise can be transmitted via common mode which could be picked up by an external measuring instrument but which wouldn't be present in the currents driving the load necessarily.
I continue to use HEXFRED and other well-regarded types of diodes because although they may not have a demonstrably audible impact on sound quality, they are excellent marketing points when I go to sell things I've built.
And if you can hear the difference, more power to you.
I replaced the garden-variety diodes in the power supply rectifier section of a Musical Fidelity XCAN V3 headphone amp with HEXFRED and afterwards measured about 1.5 dB less noise in the output of the amplifier into a 150 ohm load. The amplifier did not sound any different to me.
I did not use A weighting or any other filtering in my measurement (probably should have, in retrospect) and I measured using equipment which was flat to 20 MHz....
I suspect that the decline in noise was equal energy per octave, but as there are more ultrasonic and RF octaves than audio octaves I think that the reduction in noise was mostly at frequencies well beyond my hearing. I should have done a spectral analysis of the residual noise.
Reducing RF spuriae from diode switching in the AC supply could have an impact on audio - many devices used for voltage or current gain such as bipolar transistors and FETs can show nonlinear behavior when fed signals like these, which could have an impact on the audio they are amplifying. COULD - it's a pretty big "IF" here.
And you'd think that none of this would make it past the filter caps, but then again noise can be transmitted via common mode which could be picked up by an external measuring instrument but which wouldn't be present in the currents driving the load necessarily.
I continue to use HEXFRED and other well-regarded types of diodes because although they may not have a demonstrably audible impact on sound quality, they are excellent marketing points when I go to sell things I've built.
And if you can hear the difference, more power to you.