Hi All,
I recall the LM3875 being rated at 50W.
However, using a sinewave input of 2V-7V, I am able to get no more than 14.5V output into an 8.5 ohm load before distortion (as seen in the sinewave output on a scope) starts to set in.
This would correspond to an output power of V^2/R, which is 24.7W, much lower than the rated 50W.
Any explanations?
My setup:
IGC-LM3875, 10k input R, 220k feedback R
300W toroidals, 22V AC, no regulation, about 31V DC
32mV DC offset, 12cmx7cm aluminium heatsink,
1000uF Panasonic FC caps, pin7 to ground, No input cap
Thanks.
Ryan
I recall the LM3875 being rated at 50W.
However, using a sinewave input of 2V-7V, I am able to get no more than 14.5V output into an 8.5 ohm load before distortion (as seen in the sinewave output on a scope) starts to set in.
This would correspond to an output power of V^2/R, which is 24.7W, much lower than the rated 50W.
Any explanations?
My setup:
IGC-LM3875, 10k input R, 220k feedback R
300W toroidals, 22V AC, no regulation, about 31V DC
32mV DC offset, 12cmx7cm aluminium heatsink,
1000uF Panasonic FC caps, pin7 to ground, No input cap
Thanks.
Ryan
First of all you gain is higher than 20 which means that you are trying, with a supply of 31 V, to get at least 40 V out with 2 V in and if possible some 140 V out with 7 V. This means that already at 2 V in you amp is clipping like mad. This seems like a good reason why you don't see what you expect. Secondly the 14.5 V corresponds to perfect sine wave RMS power of (14.5*14.5)/(2*8.5) which is 12.35 W and not 24.7. Third the power rating into 8.5 Ohms with a full 31 V per rail is not 50 W but some 42 W.
Adding more caps to the supply will perhaps make it burn to pieces more easily.
Adding more caps to the supply will perhaps make it burn to pieces more easily.
Interesting answers thanks,
CARLOSFM: Adding more caps not good for the sound.
URSV: I guess you're assuming that the maximum output voltage should correspond to the rail voltage? I'll have to go bruch up on my understanding of amps.
GREGGC: Sinewave of 1kHz, did not measure rail voltage when distortion starts. I would have thought that 300 VA transformers are more than adequate and my heat sink also seems large enough, since the chip did not start to fry.
Thanks again
Ryan
CARLOSFM: Adding more caps not good for the sound.
URSV: I guess you're assuming that the maximum output voltage should correspond to the rail voltage? I'll have to go bruch up on my understanding of amps.
GREGGC: Sinewave of 1kHz, did not measure rail voltage when distortion starts. I would have thought that 300 VA transformers are more than adequate and my heat sink also seems large enough, since the chip did not start to fry.
Thanks again
Ryan
CARLOSFM: Mmm, another interesting answer. No I have not, I've relied on my fellow DIYers.
I would admit that since I have soldered the caps directly to the pins of the LM3875, it is a bit crowded and therefore any new caps will have to be added at a greater distance from the chip than the original caps.
I would admit that since I have soldered the caps directly to the pins of the LM3875, it is a bit crowded and therefore any new caps will have to be added at a greater distance from the chip than the original caps.
Dr.H said:
Your problem is not the caps. It should be clear from Usvr post what the problem is. You cannot get more than 14.5v peak from an amp with 14.5 V peak supply. Infact, because the amp needs some voltage for itself, the max output will most probably be 12v or so with 14.5 v supply.
What you need (if you want more power in 8.5 Ohms) is more supply voltage. No amount of cap upgrade of heavier xformers will change this basic fact.
CarlosFM, I'm sure you know all this. Don't put him on the wrong foot.
Jan Didden
Konnichiwa,
Yes, and that with a powersupply of +/-35V (with zero dropout at higher load - eg. regulated) and a chip temperature of 25 Centigrad, implying a humungous heatsink.
As one would expect from an Amplifier with a "soft" Powersupply and a possibly too small heatsink which would trigger the protection circuitry of the chip on sinewave testing. Note that the original gaincard is also rated as 25W RMS into 8 Ohm.
If you use the transformer you mention using one channel driven the PSU Voltage will drop to a point where the lowest points in the Supply line is around 20V. The LM3875 has around 3V dropout (clipping) voltage at that point, giving theoretically undistorted peaks of sinewaves of around 16V and thus 11.3V RMS max and 16W RMS (theoretcially).
In reality the figures may very well be a little worse or better, depending upon implementation, rectifier, cpacitor quality etc, so 25W ballpark when sinewavetesting is just fine for what you got. More interesting however, for as long as 10mS the Amplifier could (again theoretically) sustain an output of nearly 30W which is twice the power provided during sinewavetesting and for 5mS up to 40W can be supplied.
So, in reality with music (which tends to contain short peaks well above the average or RMS level) will be able to play as loud as would possible with an amplifie rhaving a "stiff" supply and around twice the outputpower when sinewave testing.
So for once the term "Music Power" actually has meaning....
Sayonara
Dr.H said:
Yes, and that with a powersupply of +/-35V (with zero dropout at higher load - eg. regulated) and a chip temperature of 25 Centigrad, implying a humungous heatsink.
Dr.H said:
As one would expect from an Amplifier with a "soft" Powersupply and a possibly too small heatsink which would trigger the protection circuitry of the chip on sinewave testing. Note that the original gaincard is also rated as 25W RMS into 8 Ohm.
If you use the transformer you mention using one channel driven the PSU Voltage will drop to a point where the lowest points in the Supply line is around 20V. The LM3875 has around 3V dropout (clipping) voltage at that point, giving theoretically undistorted peaks of sinewaves of around 16V and thus 11.3V RMS max and 16W RMS (theoretcially).
In reality the figures may very well be a little worse or better, depending upon implementation, rectifier, cpacitor quality etc, so 25W ballpark when sinewavetesting is just fine for what you got. More interesting however, for as long as 10mS the Amplifier could (again theoretically) sustain an output of nearly 30W which is twice the power provided during sinewavetesting and for 5mS up to 40W can be supplied.
So, in reality with music (which tends to contain short peaks well above the average or RMS level) will be able to play as loud as would possible with an amplifie rhaving a "stiff" supply and around twice the outputpower when sinewave testing.
So for once the term "Music Power" actually has meaning....
Sayonara
I suggest that if you have a dual trace scope, connect the other channel to monitor one of the power supply rails. This way you can literally see whether your power supply is up to the task.
You transformer is more than big enough, in fact it is overkill. But with only 1000 uF of cap on the power supply, you are asking for problems. A 40 W RMS amp (for example) will draw an average current from each supply rail of over 1 ampere (P=I^2*R, and remember each rail supplies power for 1/2 the time). Drawing 1 ampere from a 1000 uF cap will discharge it in much less than the 8.3 ms that it must hold its charge (assuming a full-wave rectifier, the power supply crests are 1/120 second = 8.3 ms apart). If you want more power, you need more capacitance.
You transformer is more than big enough, in fact it is overkill. But with only 1000 uF of cap on the power supply, you are asking for problems. A 40 W RMS amp (for example) will draw an average current from each supply rail of over 1 ampere (P=I^2*R, and remember each rail supplies power for 1/2 the time). Drawing 1 ampere from a 1000 uF cap will discharge it in much less than the 8.3 ms that it must hold its charge (assuming a full-wave rectifier, the power supply crests are 1/120 second = 8.3 ms apart). If you want more power, you need more capacitance.
> the maximum output voltage should correspond to the rail voltage?
In an ideal amplifier of this type, yes. The peak output voltage is equal-to or less-than the rail voltage (and in real life, always less-than).
Assuming you have +/-31 volts and that does not change with load, and assuming the amplifier had zero loss, you should clip at 31V peak or about 58 watts in 8Ω.
If you only had say +/-15V, perfect regulation, perfect amplifier, 15V peak is 10.5V RMS or 13.7 watts. Power goes by square of the rail voltage, so modest changes in voltage make significant changes in power numbers. (Not so much difference in Loudness: loudness is more like voltage than power, and double-voltage is not-quite double-loud.)
Real amplifiers have real losses, often 5 volts or more per side at high current. Still assuming a perfect power supply, but 5V loss in the amplifier, 31-5=26V, 26V peak is 18V RMS is 40 watts RMS.
You have an oversized transformer but undersized capacitors. Some say 1,000µFd is a magic value for "sound"; in which case don't question "authority" or ask about power (after all, who cares about power, it is sound we want to hear).
If you must do the numbers, I will guess that under load your +/-31V sags to +/-25V DC or lower at the ripple-dip. 25V-5V=20V peak, which is 14V RMS, about what you measured. 31V-25V droop is quite reasonable for an under-capped power supply.
On purely mechanical grounds, I want to see over 2,000µFd across the rails of an 8Ω amplifier eating 60Hz power. That would be two 4,000µFd caps for split supply. And at modern cap prices, I would use 2 to 5 times that much. The rails will be stable, and there won't be much buzz-breakthrough in clipping.
Other people with better ears feel that a too-stable supply is un-musical, and use 1,000µFd caps max. I see the argument for guitar amplifiers, where clipping and power-sag is part of the tone. I am unconvinced about "minimal" rail capacitance for general audio; I do know many happy designs with startlingly small power caps. I do know that in the days when I tinkered, I often liked the sound better with huge caps (and they certainly look better on the scope).
If you want big power numbers, put in big caps.
If you want "sound", try the classic under-capped design and compare with some fat caps added upstream of the chip (close to the rectifier DC output).
In an ideal amplifier of this type, yes. The peak output voltage is equal-to or less-than the rail voltage (and in real life, always less-than).
Assuming you have +/-31 volts and that does not change with load, and assuming the amplifier had zero loss, you should clip at 31V peak or about 58 watts in 8Ω.
If you only had say +/-15V, perfect regulation, perfect amplifier, 15V peak is 10.5V RMS or 13.7 watts. Power goes by square of the rail voltage, so modest changes in voltage make significant changes in power numbers. (Not so much difference in Loudness: loudness is more like voltage than power, and double-voltage is not-quite double-loud.)
Real amplifiers have real losses, often 5 volts or more per side at high current. Still assuming a perfect power supply, but 5V loss in the amplifier, 31-5=26V, 26V peak is 18V RMS is 40 watts RMS.
You have an oversized transformer but undersized capacitors. Some say 1,000µFd is a magic value for "sound"; in which case don't question "authority" or ask about power (after all, who cares about power, it is sound we want to hear).
If you must do the numbers, I will guess that under load your +/-31V sags to +/-25V DC or lower at the ripple-dip. 25V-5V=20V peak, which is 14V RMS, about what you measured. 31V-25V droop is quite reasonable for an under-capped power supply.
On purely mechanical grounds, I want to see over 2,000µFd across the rails of an 8Ω amplifier eating 60Hz power. That would be two 4,000µFd caps for split supply. And at modern cap prices, I would use 2 to 5 times that much. The rails will be stable, and there won't be much buzz-breakthrough in clipping.
Other people with better ears feel that a too-stable supply is un-musical, and use 1,000µFd caps max. I see the argument for guitar amplifiers, where clipping and power-sag is part of the tone. I am unconvinced about "minimal" rail capacitance for general audio; I do know many happy designs with startlingly small power caps. I do know that in the days when I tinkered, I often liked the sound better with huge caps (and they certainly look better on the scope).
If you want big power numbers, put in big caps.
If you want "sound", try the classic under-capped design and compare with some fat caps added upstream of the chip (close to the rectifier DC output).
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.