i don't think it's a given per se. you can't compare the two, really. the ucd is a plug-n-play amp, a clone you have to build yourself.
the ucd delivers 9,5A continously whereas a single lm3875 delivers way less.
I can only say that my ucd based amp sounds way better than every clone i've built. a linn klimax is a chipamp too.
the ucd delivers 9,5A continously whereas a single lm3875 delivers way less.
I can only say that my ucd based amp sounds way better than every clone i've built. a linn klimax is a chipamp too.
After looking through the "UCD180 questions" thread(18 BIG pages...), I can't really say who would be able to tell the story by it's name(i.e. what is better and why...)
And let's not forget, better is relative, and personal taste can influence the result a lot.
PS: bug in vBB, see the timestamp of the two posts
And let's not forget, better is relative, and personal taste can influence the result a lot.
PS: bug in vBB, see the timestamp of the two posts
matjans said:the ucd delivers 9,5A continously whereas a single lm3875 delivers way less.
I can only say that my ucd based amp sounds way better than every clone i've built. a linn klimax is a chipamp too.
Have you ever tried an LM3886 with regulated PSU?
The LM3875 is not suited to drive most modern speakers, just forget it.
I am not Carlos, but to answer your questions:
R3 and R4 are in series with a small capacitor, these can not pass DC which is what is stored inthe large filter caps. They do provide a very low impedance to high frequencies, as such will short out high frequency transients and suppress high frequency oscillation.
R1 and R2 are however a conducting path across the storage capacitors, and will in a short time drain the caps after shut down.
For Carlos: I am glad to see you are pursuing this area. But the problem can not be simply the higher inductance of larger caps as adding more caps to existing caps will lower the inductance. as inductors in parallel become smaller not larger.
R3 and R4 are in series with a small capacitor, these can not pass DC which is what is stored inthe large filter caps. They do provide a very low impedance to high frequencies, as such will short out high frequency transients and suppress high frequency oscillation.
R1 and R2 are however a conducting path across the storage capacitors, and will in a short time drain the caps after shut down.
For Carlos: I am glad to see you are pursuing this area. But the problem can not be simply the higher inductance of larger caps as adding more caps to existing caps will lower the inductance. as inductors in parallel become smaller not larger.
SheldonD said:
I've been out, thanks Sheldon.
SheldonD said:
Maby there are other reasons too, I'm open to suggestions.
Anyway, on the chip there are 100uf caps bypassed with 100nf ceramics.
I've made many amps with several chips, several implementations, and tried too many combinations regarding unregulated PSU.
2x3,300uf per rail sound worse than single 4,700uf caps per rail.
Really.
2x4700uf sound as awful as 10,000uf.
Go lower to 1,000~1,500uf per rail/chip and that's the sweet spot.
But then most probably it won't drive your speakers.
This snubber lets me go higher in capacitance without mucking up the sound.
Thanks alll for your answers.
I thought the discharge resistors had to be in front of the capacitors like on Dejan's schematics, which is why i asked.
Anyway, i have more questions:
1-Carlos, why didn't you implement the snub capacitors across each one of the bridge diodes like on Dejan's schematic?
I know Nuuk also recommends doing this on his web pages...
2-How come nobody here building PSUs recommends using at least one termistor (Digikey Part#KC006L-ND) to limit the Inrush current in the PSU?
From what i have read, it has many the benefits..
I know Nelson Pass has termistors in at least one of his PSU designs
Thanks
I thought the discharge resistors had to be in front of the capacitors like on Dejan's schematics, which is why i asked.
Anyway, i have more questions:
1-Carlos, why didn't you implement the snub capacitors across each one of the bridge diodes like on Dejan's schematic?
I know Nuuk also recommends doing this on his web pages...
2-How come nobody here building PSUs recommends using at least one termistor (Digikey Part#KC006L-ND) to limit the Inrush current in the PSU?
From what i have read, it has many the benefits..
I know Nelson Pass has termistors in at least one of his PSU designs
Thanks
Having a little trouble understanding why the R in the snubber has to be such high wattage. Regardless of the interaction with filter capacitor inductance, isn't it still just "snubbing" high frequency? In the case of Carlos' schematic with 1R/0.120uf snubber the f > 1MHz. Can there be this much energy at high frequency in the PSU or is there something else going on?
Veselinovic was using up to 17 watt R's for 50 to 100 watt amps(me thinks). Did some digging. The original design of the Otala inspired amp that Veselinovic based his usage of the RC snubber on used 1 watt R's:
http://home.online.no/~tsandstr/Otala amp 1973, original.htm
and the production design built by Electrocompaniet used 0.5 watt R's for the snubber:
http://home.online.no/~tsandstr/schematics_of_the_real_ec_25w_am.htm
This was a 25 watt amp.
I understand a little wattage headroom will be a good idea for a stable filter but shouldn't two watts be enough for a 30 to 50 watt GC?
And thanks Carlos for pointing out this filter.
scottw
Veselinovic was using up to 17 watt R's for 50 to 100 watt amps(me thinks). Did some digging. The original design of the Otala inspired amp that Veselinovic based his usage of the RC snubber on used 1 watt R's:
http://home.online.no/~tsandstr/Otala amp 1973, original.htm
and the production design built by Electrocompaniet used 0.5 watt R's for the snubber:
http://home.online.no/~tsandstr/schematics_of_the_real_ec_25w_am.htm
This was a 25 watt amp.
I understand a little wattage headroom will be a good idea for a stable filter but shouldn't two watts be enough for a 30 to 50 watt GC?
And thanks Carlos for pointing out this filter.
scottw
the snubber's job is to damp the ringing of the diodes as they switch on and off -- the frequency of the ringing is a lot higher than the audio spectrum -- but this also means that the snubber itself should be placed as closely as possible to the rectifiers -- any trace distance adds inductance to the circuit -- and this changes the impedance of the supply -- of course this is a problem of enormous import to switchers, a lot less so for linear supplies .... perhaps tomorrow I will import the SPICE models for the MUR860 ...carlosfm said:
yuri777 said:
Everytime I take those caps across diodes on a CDP I have improvements.
So, I don't use that approach on my amps.
scottw said:
My schematic is preliminary, It is based on my test.
I also think that Dejan quite exagerated on the power of the snubber resistors.
Maby 2~3W is enough.
ok, here's the math --
somehow my last post disappeared when I hit enter, i hope i am not being repetitive.
you have to look at the snubbers as part of the network made up of the power supply elements and the amplifier. it doesn't matter that you have two of them (100nF and the RC of 120nF and 1R) as was drawn.
you need to know at least 2 things, and probably 3: 1) the inductance of the transformer secondary (with the primary shorted) [br] 2) the combined capacitance of the diode AND transformer primary to secondary, and [br] the peak to peak "repetitive" voltage.
for one of my 60Hz torroids I measured the secondary leakage inductance as 180 uH, the MUR860 has capacitance of 100pF at 30V, and the transformer capacitance is 300pF. The pk to pk Vrrm maybe 100mV.
the resonant frequency is F= 1/(2pi sqrt LC) = 593kHz
the snubber resistor is Rs = sqrt (L/C) = 670R
the snubber capacitor is Cs = 1/(Rs * F) = 2.5 nF
the resistive dissipation is P = 1/2 * Cs * (Vrrm^2) * F = 7.4 uW
the math is a little easier on ON-Semi's website than Hagerman's paper. http://www.onsemi.com/pub/Collateral/HB214-D.PDF -- look at page 170.
if you use much larger capacitors the power dissipation in the snubber resistor is going up a lot, but not enough to spoil a 1/2 watt resistor -- in a switcher it becomes problematic. It would appear that a larger capacitor will add more damping and reduce the ill-effects of diode-ringing.
somehow my last post disappeared when I hit enter, i hope i am not being repetitive.
you have to look at the snubbers as part of the network made up of the power supply elements and the amplifier. it doesn't matter that you have two of them (100nF and the RC of 120nF and 1R) as was drawn.
you need to know at least 2 things, and probably 3: 1) the inductance of the transformer secondary (with the primary shorted) [br] 2) the combined capacitance of the diode AND transformer primary to secondary, and [br] the peak to peak "repetitive" voltage.
for one of my 60Hz torroids I measured the secondary leakage inductance as 180 uH, the MUR860 has capacitance of 100pF at 30V, and the transformer capacitance is 300pF. The pk to pk Vrrm maybe 100mV.
the resonant frequency is F= 1/(2pi sqrt LC) = 593kHz
the snubber resistor is Rs = sqrt (L/C) = 670R
the snubber capacitor is Cs = 1/(Rs * F) = 2.5 nF
the resistive dissipation is P = 1/2 * Cs * (Vrrm^2) * F = 7.4 uW
the math is a little easier on ON-Semi's website than Hagerman's paper. http://www.onsemi.com/pub/Collateral/HB214-D.PDF -- look at page 170.
if you use much larger capacitors the power dissipation in the snubber resistor is going up a lot, but not enough to spoil a 1/2 watt resistor -- in a switcher it becomes problematic. It would appear that a larger capacitor will add more damping and reduce the ill-effects of diode-ringing.
There's some confusion here it seems.
The snubbers in the original schematic are going to be innefective at damping the transformer resonance, since they are the wrong side of the rectifiers and way too far away to be of any practical use, in this respect.
Any benefits Carlos is getting are much more likely to be related to the decoupling to the IC itself.
As an addition though, the 100n caps in parallel with the transformer don't seem like a good idea either. Based on the fact that the transformer will have a resonance, these just add more C and lower the resonant frequency closer to the audio band, solving one problem (RF noise) and making another (AF noise). It may be a better net result if the reservoir caps are lower Z at this frequency, but it's far from ideal. Better to snub the oscillation properly (which requires R not C), the C is there solely to reduce Pd in the resistor - it is, in effect, an AC-coupled resistor.
If you use the figures presented by jackinj as an example of a typical transformer, you move it's self-resonance from 600kHz to around 37kHz by adding that 100n.
It shouldn't have any significant effect, if the C value / type is chosen correctly (i.e. is low Z at the resonant frequency). The R provides the damping.
It's not the diodes ringing either...
Andy.
The snubbers in the original schematic are going to be innefective at damping the transformer resonance, since they are the wrong side of the rectifiers and way too far away to be of any practical use, in this respect.
Any benefits Carlos is getting are much more likely to be related to the decoupling to the IC itself.
As an addition though, the 100n caps in parallel with the transformer don't seem like a good idea either. Based on the fact that the transformer will have a resonance, these just add more C and lower the resonant frequency closer to the audio band, solving one problem (RF noise) and making another (AF noise). It may be a better net result if the reservoir caps are lower Z at this frequency, but it's far from ideal. Better to snub the oscillation properly (which requires R not C), the C is there solely to reduce Pd in the resistor - it is, in effect, an AC-coupled resistor.
If you use the figures presented by jackinj as an example of a typical transformer, you move it's self-resonance from 600kHz to around 37kHz by adding that 100n.
It shouldn't have any significant effect, if the C value / type is chosen correctly (i.e. is low Z at the resonant frequency). The R provides the damping.
It's not the diodes ringing either...
Andy.
i'm sure we've all had this problem - the circuit is working OK, but not great (distortion is high, etc.) and you put your finger on an IC -- WOW is it HOT !!! -- the problem with these resonant circuits is that the oscillations find their way to where they shouldn't be! the IC gets overdriven, etc., etc.
not that 600kHz is a particularly short wavelength, but you know what I mean.
and with a GC chip we know to be mindful...
Hagerman sources an article in Audio Amateur 3/94 which discusses snubbers.
not that 600kHz is a particularly short wavelength, but you know what I mean.
and with a GC chip we know to be mindful...
Hagerman sources an article in Audio Amateur 3/94 which discusses snubbers.
ALW said:
You and jackinnj are missing the point here.
I can understand that caps on the primary may not always be good, ok.
But the snubber after the big caps have nothing to do with the diodes.
Let me explain how I made the tests:
1. LM4780 PCB with 100uf+100nf on-board.
2. PSU with (basically) 3,300uf caps per rail.
4. Listen - sounds fine, but my speakers ask for tighter bass
5. Add another 10,000uf cap per rail on the PSU
6. Listen - bass ok, but sounds bad (midband and treble like a cheap amp)
7. Add the snubber (1R + 560nf) - better
8. After testing several snubbers, final 1R+120nf -
So, the chip is the same, the bypassing on the chip too.
What made those big 10,000uf caps sound good with this chip was the snubber.
13,300uf of capacitance per rail with and unregulated PSU and with a chip amp never sounded so good to me.
jackinnj said:
The LM4780 is palyin' on my main system for almost a week, with my tough EPOS ES11 speakers.
Heatsing is a black anodized P3 CPU cooler, without isolator on the chip.
Actually, this works very well, between the chip and the heatsink there's only thermal paste, but the chip is isolated.
I've pushed it hard and the heatsink doesn't get hot to touch, and in normal listening levels it doesn't even get warm.
I would not put this on my main system for a test if it got too hot as you say.
On my bench it had been playing for some days, with 4 ohm test speakers.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.