Master's words
Just found in the article regarding regulated power supply for Zen on PassDiy.com
These are the words of the Master, do you believe me now?😉 😉
Cheers
Andrea
Just found in the article regarding regulated power supply for Zen on PassDiy.com
These are the words of the Master, do you believe me now?😉 😉
Cheers
Andrea
Andypairo, I always believe you.
I can explain my point of view:
1. Heatsinks are the main issue here. IMO 0.35 deg/W won’t give you about 110W dissipation at reasionable temperature in the real world. Therefore lowering the bias will be a must. About to 4,5 – 5A total bias, I think.
2. Due to hum issue, it will probably be nessesary to go with CRC or CLC filter on supply rails. This will decrease supply voltage to about 15 – 16V.
3. And what we get than? The standard Grey’s version !
Best regards😉
I can explain my point of view:
1. Heatsinks are the main issue here. IMO 0.35 deg/W won’t give you about 110W dissipation at reasionable temperature in the real world. Therefore lowering the bias will be a must. About to 4,5 – 5A total bias, I think.
2. Due to hum issue, it will probably be nessesary to go with CRC or CLC filter on supply rails. This will decrease supply voltage to about 15 – 16V.
3. And what we get than? The standard Grey’s version !
Best regards😉
I am acquiring a circuit that has 4 fets per channel. I have recalculated that using 16V and a 5A bias creates 160W dissipation. That is 40W per Fet (so 80W per heatsink). That is 28.8degsC into a 0.36degsC/W (sorry my fault - not 0.35 as previously stated)...That means a heatsink temp of 49.8degsC (or in reality, about 55degsC).
How's that then? It's still 49W into my 8R speakers.
Thanks,
Gaz
How's that then? It's still 49W into my 8R speakers.
Thanks,
Gaz
Hi
If you don’t want to build the monster and costs are important, it’ll be good choice. IMO.
With my experiences and others, calculations on power of AX are too optimistic, particularly for low rail voltage. With 16V rail you can get about 20V peak of signal swing , than maksimum power will be 50W peak, and 35W rms.
Needed current for this is 2.5A (per side) and 5 A for total. Remember, that AX has active current sources with 50% current gain, therefore we don’t need 5A for total bias. This is your margin of possibility to decreasing bias lower than 5A (theoreticaly to 2.5A). But from the other side more bias - less distortions and better sound. We discuss 8 ohm speakers of course.
Regards
If you don’t want to build the monster and costs are important, it’ll be good choice. IMO.
With my experiences and others, calculations on power of AX are too optimistic, particularly for low rail voltage. With 16V rail you can get about 20V peak of signal swing , than maksimum power will be 50W peak, and 35W rms.
Needed current for this is 2.5A (per side) and 5 A for total. Remember, that AX has active current sources with 50% current gain, therefore we don’t need 5A for total bias. This is your margin of possibility to decreasing bias lower than 5A (theoreticaly to 2.5A). But from the other side more bias - less distortions and better sound. We discuss 8 ohm speakers of course.
Regards
Let's get math
Jarek, I agree that with slightly under-dimensioned heatsinks lowering the bias is the most effective solution, but what I wanted to point out is that, given a heatsink temp (which is function of total power dissipation), two fets will run (much) cooler than one:
Some calculations using Gaz's situation:
40W / fet
55 °C heatsink temp
Looking at the datasheet if IRPF240 thermal resistance junction to case is 0.83 °C/W, you must add from 0.3 to 1 or more for insulators, so we have 1.13 °C/W from junction to sink in the best situation.
This yelds a junction temp (the only value that really matters) of 100 °C (128.5 °C); the worse situation is too close IMHO to the max junction temp, which is 150°C.
With 2 fets the temperature rise between heatsink and junction would be halved, giving respectively 77.5 °C and 91.5 °C, still acceptable.
In this thread also is pointed out the importance of a good insulator: http://www.diyaudio.com/forums/show...d=2863&highlight=insulator+thermal+resistance
Compared to what was said there I have even been a little optimistic 🙂
So if the thermal resistance of insulators is unknown I wouldn't risk.
Cheers
Andrea
Jarek, I agree that with slightly under-dimensioned heatsinks lowering the bias is the most effective solution, but what I wanted to point out is that, given a heatsink temp (which is function of total power dissipation), two fets will run (much) cooler than one:
Some calculations using Gaz's situation:
40W / fet
55 °C heatsink temp
Looking at the datasheet if IRPF240 thermal resistance junction to case is 0.83 °C/W, you must add from 0.3 to 1 or more for insulators, so we have 1.13 °C/W from junction to sink in the best situation.
This yelds a junction temp (the only value that really matters) of 100 °C (128.5 °C); the worse situation is too close IMHO to the max junction temp, which is 150°C.
With 2 fets the temperature rise between heatsink and junction would be halved, giving respectively 77.5 °C and 91.5 °C, still acceptable.
In this thread also is pointed out the importance of a good insulator: http://www.diyaudio.com/forums/show...d=2863&highlight=insulator+thermal+resistance
Compared to what was said there I have even been a little optimistic 🙂
So if the thermal resistance of insulators is unknown I wouldn't risk.
Cheers
Andrea
Andypairo,
Your calculations are correct for me. But I spoke about IRFP044’s which have up to 175 deg of Celsius operating junction temperature. To be more safety I would decrease power dissipation to 35W per Fet by lowering bias current.
Regards
Your calculations are correct for me. But I spoke about IRFP044’s which have up to 175 deg of Celsius operating junction temperature. To be more safety I would decrease power dissipation to 35W per Fet by lowering bias current.
Regards
Hi,
OK, here goes...
I modelled my PSU and the sim says it will produce 16.32V (I don't expect it to be that precise - but I'll go with it for now).
As I said before - the circuit board has room for four FETs and I will use IRFP044 (I have found a source!).
For the following setup, 4.3A of bias will produce almost spot on 35W dis per FET (or 70W per heatsink).
Does that sound OK? Now - a couple of questions. The bias is for the whole amp (monoblock) - so each fet should have 0.93A of bias, which I measure over the resistor for voltage and use Ohms law to calculate. Is this correct?!
Thanks,
Gaz
OK, here goes...
I modelled my PSU and the sim says it will produce 16.32V (I don't expect it to be that precise - but I'll go with it for now).
As I said before - the circuit board has room for four FETs and I will use IRFP044 (I have found a source!).
For the following setup, 4.3A of bias will produce almost spot on 35W dis per FET (or 70W per heatsink).
Does that sound OK? Now - a couple of questions. The bias is for the whole amp (monoblock) - so each fet should have 0.93A of bias, which I measure over the resistor for voltage and use Ohms law to calculate. Is this correct?!
Thanks,
Gaz
Rarkov said:
in my case The rail voltage turned out to be lower by 1 volt or so with respect to the sims. I thought that was because everything gets really hot and simulation do not accnt for thermal effects, just my guess.
am I right is assuming you mean a total of 4 mosftets/chn (no parallel mosfets)? under those circumstances it will be 4.3/2 amps or 2.15 per fet to be measured across either source resistor. In any event, by design the drop should be kept to .5 volts so calculate the proper source resistor.
You are probably going to be limited by the type and values of available power resistors.
Hope that helps.
You're correct - no parallel FETs.
I don't think I can plan much more until I actually get the circuit here...I'll let you know!
Thanks,
Gaz
I don't think I can plan much more until I actually get the circuit here...I'll let you know!
Thanks,
Gaz
Nelson Pass
I have been looking through this thread and was not able to find an answer to this question. Did you address it in another thread? If so would it be possible to direct me to it? From what I have read and surmised, we need to bring the rail voltage high enough so that it will act on the best (most linear) part of the FET's output curve. Is this correct? If so, the derating curve will be dependent on the FET used so what would be the optimum rail voltage to use with the FETs below for the Aleph-X in your experience?
IRFP044
IRFP044N
IRFP150N
IRFP240
The other issue is the bias. There are posts where you state outright that more will always sound better but is this true if a speaker driver's impedence is high enough to need no additional current even at its worst dip? Is there a point in the speaker impedence/amp relationship where raising the voltage rail is more effective than adding more current bias?
Thanks in advance for your reply.
I have been looking through this thread and was not able to find an answer to this question. Did you address it in another thread? If so would it be possible to direct me to it? From what I have read and surmised, we need to bring the rail voltage high enough so that it will act on the best (most linear) part of the FET's output curve. Is this correct? If so, the derating curve will be dependent on the FET used so what would be the optimum rail voltage to use with the FETs below for the Aleph-X in your experience?
IRFP044
IRFP044N
IRFP150N
IRFP240
The other issue is the bias. There are posts where you state outright that more will always sound better but is this true if a speaker driver's impedence is high enough to need no additional current even at its worst dip? Is there a point in the speaker impedence/amp relationship where raising the voltage rail is more effective than adding more current bias?
Thanks in advance for your reply.
Very good questions.
In general, after you get more than 2 volts or so Drain to
Source, the device pretty much acts like a voltage controlled
current source. As the voltage increases, this tends to be
even more the case, with a higher transconductance number.
The linearity of the transistor has similar behaviours in the
case of bias and voltage. First off, the transconductance
increases with current (up to a rather high point, anyway)
and so you find that the distortion due to variations in current
decreases, but of course it is also true that your signal current
will also likely be a smaller proportion of the bias current. Both
these effects will lower the distortion, keeping in mind that
higher chip temperature does lower the transconductance.
With voltage, the transconductance increases with voltage, and
all the junction capacitances will decrease with increased
Drain-Source voltage (this is good for bandwidth), and once
again, the when the output signal swing is small compared
the DC "bias", the distortion is less.
So basically this means that for maximum linearity, we want to
run our devices at high current and high bias. Unfortunately
the dissipation involved can cut our activity short, so it always
comes down to some sort of compromise.
It's different for each device, but a little experimentation can get
you the information you want. In the case of the newly published
ZV5, we see that for a given load impedance and desired
wattage, there is a bias voltage and current beyond which you
get diminishing returns. Check out Figures 4, 5, and 6 of that
article at www.passdiy.com
In general, after you get more than 2 volts or so Drain to
Source, the device pretty much acts like a voltage controlled
current source. As the voltage increases, this tends to be
even more the case, with a higher transconductance number.
The linearity of the transistor has similar behaviours in the
case of bias and voltage. First off, the transconductance
increases with current (up to a rather high point, anyway)
and so you find that the distortion due to variations in current
decreases, but of course it is also true that your signal current
will also likely be a smaller proportion of the bias current. Both
these effects will lower the distortion, keeping in mind that
higher chip temperature does lower the transconductance.
With voltage, the transconductance increases with voltage, and
all the junction capacitances will decrease with increased
Drain-Source voltage (this is good for bandwidth), and once
again, the when the output signal swing is small compared
the DC "bias", the distortion is less.
So basically this means that for maximum linearity, we want to
run our devices at high current and high bias. Unfortunately
the dissipation involved can cut our activity short, so it always
comes down to some sort of compromise.
It's different for each device, but a little experimentation can get
you the information you want. In the case of the newly published
ZV5, we see that for a given load impedance and desired
wattage, there is a bias voltage and current beyond which you
get diminishing returns. Check out Figures 4, 5, and 6 of that
article at www.passdiy.com
Guys,
I'm building the original mini-XA. Which power mosfet do I need to match?
Q1 & 2 and Q10 & 11 or all 4 should be equally matched? Or, none of them?
Note: I searched the threads. Was getting headache searching through all them.
thanks.
Vince
I'm building the original mini-XA. Which power mosfet do I need to match?
Q1 & 2 and Q10 & 11 or all 4 should be equally matched? Or, none of them?
Note: I searched the threads. Was getting headache searching through all them.
thanks.
Vince
Match the input diff pair, then ideally match all the
transistors on the negative rail (both sides). On the
positive rail you need only match the devices that
are parallel, not both sides to each other.
transistors on the negative rail (both sides). On the
positive rail you need only match the devices that
are parallel, not both sides to each other.
Nelson Pass
Earlier in this thread you had mentioned that you would post a test circuit/jig that would help us do this beyond matching Vgs. I have been unable to locate it. If you posted it in another thread could you point me to it?
You always give us more than we ask but in this case I think I would have preferred less explanatory replies based on your extensive experience with these FETs. If I use the IRFP150N FETs in the Aleph-X circuit described in this thread, what voltage rail would put me into the most linear part of its output curve? Would there be an audable change if I lowered the bias through each of these FETs from 2A to 1A? What if the speaker impedence were high enough so that it had sufficient current at 1A of bias? Assuming my heatsinks can dissipate 0.25C°/W how many watts can I put through this FET before seeing/hearing the aforementioned transconductance non-linearities from heat.
yldouright,
I've recently been in contact with Nelson regarding this issue and he tells me that we should expect details of the test-rig sometime this week.
I've recently been in contact with Nelson regarding this issue and he tells me that we should expect details of the test-rig sometime this week.
Slight delay. Had to pack my son off for his first year
at UCSB, got the horrible hacking bug he left behind, and
this morning we had to put down my wife's 31 year old
champion horse.
It won't happen till next week, sorry.
at UCSB, got the horrible hacking bug he left behind, and
this morning we had to put down my wife's 31 year old
champion horse.
It won't happen till next week, sorry.
Hi Nelson,
Not quite what is was you're son left behind but 31 years is a very exceptional age for any horse...
Guess he/she was a thoroughbred...
Cheers,😉
Not quite what is was you're son left behind but 31 years is a very exceptional age for any horse...
Guess he/she was a thoroughbred...
Cheers,😉
Very exceptional horse, from the Raffles blood line. My wife
Jill rode him to 11th on the 100 mile Tevis, and he was the
best condition horse across the line. That was 20 years ago,
however.
Jill rode him to 11th on the 100 mile Tevis, and he was the
best condition horse across the line. That was 20 years ago,
however.
To Jill and Nelson.
A MILLION HORSES
Don't cry for the horses
That life has set free
A million horses
Forever to be
Don't cry for the horses
Now in God's hands
As they dance and they prance
To a heavenly band
They were ours as a gift
But never to keep
As they close their eyes
Forever to sleep
Their spirits unbound
On silver wings they fly
A million horses
Against the blue sky
Look up into heaven
You'll see them above
The horses we lost
The horses we loved
Manes and tails flowing
They Gallop through time
They were never yours
They were never mine
Don't cry for the horses
They will be back someday
When our time has come
They will show us the way
On silver wings they will lift us
To the warmth of the sun
When our life is over
And eternity has begun
We will jump the sun
And dance over the moon
A Ballet of horses and riders
On the winds to a heavenly tune
Do you hear that soft nicker
Close to your ear?
Don't cry for the horses
Love the ones that are here
~ Author Unknown ~
/Hugo
A MILLION HORSES
Don't cry for the horses
That life has set free
A million horses
Forever to be
Don't cry for the horses
Now in God's hands
As they dance and they prance
To a heavenly band
They were ours as a gift
But never to keep
As they close their eyes
Forever to sleep
Their spirits unbound
On silver wings they fly
A million horses
Against the blue sky
Look up into heaven
You'll see them above
The horses we lost
The horses we loved
Manes and tails flowing
They Gallop through time
They were never yours
They were never mine
Don't cry for the horses
They will be back someday
When our time has come
They will show us the way
On silver wings they will lift us
To the warmth of the sun
When our life is over
And eternity has begun
We will jump the sun
And dance over the moon
A Ballet of horses and riders
On the winds to a heavenly tune
Do you hear that soft nicker
Close to your ear?
Don't cry for the horses
Love the ones that are here
~ Author Unknown ~
/Hugo
Hi,
Hugo, that's a very touching tribute.
At my parents' we had horses of all kinds...
The most extraordinary ones were the Arab thouroughbreds of which we had a Polish-Arab (Gazal VII) mare who reached the age of 33 before passing away and a magnificent Eastern-Arab stallion called Orlando, a son of Ghibli.
To me, and no doubt many others, horses are the most beautiful and intelligent creatures God has put on this planet...they're just a feast to the eye.
Sorry for the OT...just love horses...😉
Hugo, that's a very touching tribute.
At my parents' we had horses of all kinds...
The most extraordinary ones were the Arab thouroughbreds of which we had a Polish-Arab (Gazal VII) mare who reached the age of 33 before passing away and a magnificent Eastern-Arab stallion called Orlando, a son of Ghibli.
To me, and no doubt many others, horses are the most beautiful and intelligent creatures God has put on this planet...they're just a feast to the eye.
Sorry for the OT...just love horses...😉