ALEPH4 - URGENT NOTICE !!!!!

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
For those who noted the error in the bias current draw on the Aleph4 schematic, please note the following reply from Nelson Pass:
_______________________________

"Well, there are a couple more errors on the schematic which resulted from it being swiped from the Aleph 2 schematic. Among them are X3 on the output stage should really be X6, for 24 output devices per channel. The voltage is
also a little low, and should be more like 6.5 volts.

Sorry about this. I will post corrections the web site. In your case, I recommend that you replace the 1.5 ohm resistors with .68 ohm devices, and otherwise leave the amplifier unmodified.

best regards, np"
________________________________

Now didn't this really depress me :-(

there are aspect of the above which don't make sense either. I SUGGEST ALL POTENTIAL BUILDERS STOP & WAIT. I will get this all sorted out and will post a complete update when I'm finished.

Cheers

Mark
 
Blmn & Others,

Conceeded!

Actually, this raises a very important question for the would-be-DIY builder. Nelson has *very conservatively* engineered these amps for commercial supply. Given the significant cost he has aimed to absolutely minimize device failure, eg. IRF244 running at 0.5A and rated at 15A.

Even the 200W version (12 devices per bank @ 4A) could well be scaled down for personnal construction where cost is an issue for people. 6 devices @ 2/3A would save many $$.

My amp becomes a "cutting-edge" test. With 3 devices @ double the idle current (ie. 1A each cf 0.5A) we still have a large margin of safety.

As blmn states, the issue will be heat dissipation. Theoretically we could go to 1 device @ 3 amps, but getting the heat out may well prove difficult. (Matching may be easier though!)

I had previously stated my amp ran "comfortably warm", which was out-of-keeping with any description of the "real thing" and we now know why!!!

I am pleased to say, it now runs hot, fitting the usual description, "this amp runs hot, you can touch the heatsinks, albeit briefly". Alas, there was no room for dC/W calculations as I got the sinks surplus.

Thanks to all, and as stated, once all this is sorted there will be a complete update file available on the Pass site.

mark
 
Aleph 5

Good to get things sorted!

I also think there is are some faults in the Aleph 5 schematics. If you look at the Volksamp A60 it states that it has 12 fet's as in the original Aleph design.

Looks like the Volksamp schematics are correct and the Aleph 4 and 5 are not.

BTW. If you look at the Zen Revisited article, Pass states that "Second, the amount of bias on the Mosfet will make an appreciable difference: Fig 2 shows an IRF140 biased at 2 amps (top curve) and 3 amps (middle curve), and the distortion has dropped by half." and also "Fourth, you can parallel the channels at the input and outputs to obtain still lower distortion figures. The bottom curves of Fig 2 and Fig 3 were obtained by parallel the channels with 3 amp bias."

Maybe it is better to run at a higher bias current, than with more paralell devices?


-Kaj
 

blmn

Member
2001-02-01 2:43 am
.
Mefinnis,

You can measure with some precision your heatsink thermal resistance. You need only a thermometer to do it. Just turn on your amp. without input signal and, after the amp reaches its normal temperature measure the temperature of the heatsink near the output devices (not over them). Use the following relation:(Tempheatsink - Tempambient)/(48W*number of devices on the heatsink).I suppose the power supply has a low internal resistance.

With only 3 devices you have, considering the ambient temperature equals 25 Celsius and the temperature on the heatsink equals 60C, 135C in the junction, near the maximum operational temperature claimed on the device's datasheet, under quiescent conditions and 0.5C/W insulator sheet thermal resistance. I think you have to analyse this question, specially if you live in a hot place, to achieve a good reliability for your amp.

About McKajVah question, considering the junction temperature, under normal conditions, safe, the minimum number of devices must consider the linear region of the devices and the designer's goal on the project. Since Mr. Pass said there was no problem with 3 devices, I think anything between 3 and 6 devices may work well.

I wish I could help you

Regards,
 
blmn,

Thanks. I'm not EE-based, so may appear a little slow. Each heatsink has 6 devices (3=current source + 3=output device) all running quiescent @ 3amps = 1A per device. I assume the 48W comes from 48VDC*1A.

Can I assume thermal R is like electrical R in that it decreases with parallel devices? Should not each FET see the thermal R of its own washer? If we have 48W/device, would not the device temp be T(d) = T(heatsink) + (Watts/device)*(thermal R of insultor):

T(d) = 60 + (48*0.5) = 84C

Otherwise, if I assume 6 devices @ 50W = 300W * 0.5C/W = 150C, then add this to ambient the poor old FET is getting just a little warm!!!!!

Appreciate this is likely very basic for you, but the help *is* appreciated.

cheers
 

blmn

Member
2001-02-01 2:43 am
.
Mefinnis,


First of all, my native language is not English, so excuse me if sometimes my texts are, how could I say, a little rude. This is not my intention, for sure, it's just a lack of good words to use. Excuse my thousand mistakes either (HI!!).

You are right, I considered the worst condition all the voltage under the devices, so the power is 48W.

In your case you can consider a good thermal resistance for the heatsink (for 60 Celsius over a external heatsink and 25 Celsius tamb) equal to (60 - 25)/(48W*6devices)= 0.12 C/W.

You can consider also the thermal resistance for each device, using this equation:

(Tempjunction - Tempheatsink)/48W = (RJunctiontocase + RInsulator).

From the IRF244 datasheet you have Rjunctiontocase=1 C/W maximum and I used .5 C/W for a insulator with thermal grease. In this case, for Tempheatsink = 60 C and the right side of the equation equals 1.5 C/W you have Tempjunction equals 48W*1.5 + 60 = 132C wich I rounded to 135.

I used .5 C/W for the insulation with some safety margin. If you know this value for your project, you will achieve lower temperature in the junction's device.

Until the heatsink, you must consider isolated thermal circuits, except if you have one big insulator sheet. With one insulator for each device you are right in your approach, you have to consider parallel thermal resistances, but each resistance in this parallel circuit is the summing of junction to case thermal resistance and insulator thermal resistance. The Td in your equation is the temperature on case of the device. To obtain the junction temperature you have to add the Junction to case thermal resistance to Rinsulator.

I was worried about the maximum junction to case thermal resistance of IRF244 Harris datasheet, because I think it was very high for a TO3 device. It was my motivation to advice you about the heatsink when Mr. Pass said you could reduce Rsource and maintain the number of the devices.

Anyway, don't worry about your non-EE mind, because your approaches on your texts are, in my opinion, very fine logical and fine.

Regards,
 
Rth = thermal resistance (seen as electric resistance)
dT = difference in temperature (seen as potential dif. or voltage)
P = thermal dissipation (seen as current)
Rth = dT / P °C/W (or °K/W same thing)
If you have 48 Watts and a heatsink rated at 0.5 °K/W
then you have dT =Rth*P = 0.5*48 = 24 °C more
This means if the ambient temp is 25 on the heatsink you
have 49.
If the mica insualtor has R = 0.5 °K/W
then we have dT = 0.5*48 = 24 °C more on the transistor case. Thats 73 °C.
If the transistor has Rjc (R junction to case) 0.7 °K/W
then we have dT = 0.7*48 = 33.6 °C
That´s 106.6 °C on the junction. Quite hot !!!
The advantage of parallel transistors is that for the same heatsink temperature we have lower junction temp.
This means we can allow a bigger heatsink temp.
For instance we have 3 trans. dissipating at NOW 48/3 = 16W.
the heatsink still has to get rid of 48 Watts so it has a temp again of 49 °C.
The insulator now makes a temp difference
dT = 0.5°K/W * 16 watts = 8 °C that´s 49 °C + 8°C = 57 °C
on the transistor case.
for the junction now we have
dT = 0.7 °K/W * 16 watts = 11.2 °C that´s
57 °C + 11.2 °C = 68.2 °C in the junction.
So each of the three transistors run cooler but they make the same heat on the heatsink.
This gives us a few choices.
1. /Smaller heatsinks, less money, hotter heatsinks (DIYers don´treally care if the heatsinks are too hot).
2. /Bigger bias current if we want lower distortion. More power in 4 ohm speakers. The heatsink runs hotter but more transistors are still cooler then one.
3. /Higher power supply voltage, more power
 
Desmond has updated the Service Manual for the Aleph4 on the passlabs website. I think we can safely assume it is now correct ;-)

My amps measured heatsink T = 66C, to which np states...

"Glad that it's worked out, after the embarassment of the incorrect schematic. The heat sinks are a bit hot for my taste, but then I try to have these things last 20-30 years before I have to repair them."

The later bit is the key to the commercial designs.

I am in the process of putting all of this information onto the web and will post the address in due course.

mark
 
Hi - A few findings of my own with heatsinking - the BIGGEST issue with Class A. I'm doing Zen Revisited - 2 sets of paralleled amps for 4ohm spkrs, each running 4amp bias, 50v rail, thats 128w on the gain stage MOSFET (for symetric clipping it runs at ~60% rail, not 50%). So MOSFET at 128w = non-trivial h/s problem. Using 250W TO3 devices (internal junc/case is 0.5C/W), which with temp. derate makes a max safe junc temp of around 120C. Yep, that's 800w total! Winter listening only?

I found some TO3 plastic washers from http://www.bergquist.com - 0.2C/W (yep, 0.TWO)(confirmed by my measurements), SILPAD K10 or SILPAD 2000 series, used dry. To get this heat out, the problem with few devices - they are mounted on to 12.5mm Aluminium plate - works a treat. Then I have a system of alternate 6*25mm Al spacers bars and more 3mm plates (the fins)- all gunged up with thermal grease (beware - not too much) and clamped together with 3 runs of 6mm studding to the 12.5mm plate.

Performance (natural convection)0.17C/W, (with fan) 0.11 C/W. Not bad - but I'd not recommend the metal bashing involved. All mating surfaces were wet rubbed with 400 grade wet/dry to ensure as flat as possible (= good thermal joint). The grease and clamping does the rest. (I asked about welding - the guy said no-go as the heatsink effect was too great - ironic really). The amp is built around the heatsink, no other chassis apart from a supporting cradle. Fan is a stripped down 30cm desk fan, run ever so slowly so it's silent (a design factor of mine - NO noise apart from music) - at around 1/2 voltage. Seems to be a good approach generally for lo/no noise fans. It still blows enough. Pics of this monster will hopefully be on NP's site in due course.

So at 22C ambient my devices are running at case temps around 56C (junc temps 56+(0.5*128) = 120, not too bad. Remember NP's comment on MOSFET service life - low temp = long life.

A 50C thermal switch on the sink protects against those hot summer days (in the UK?) or fan failure.

Hope this is useful to someone.

Pete
 
BeO washers

Nice

Have you checked out BeO washers for electrical isolation?

That is the absolute best you can get unless you bolt on to a piece of copper and use a larger area washer.

Another cool thing about BeO washers is that they are thick and so reduce capacitances of device to heatsink ....

The only problem with BeO is that they may be illegal in the EU and that if sanded they produce dust which gets you berylliosis, similar to astbestosis from asbestos ...

I have worked with this material in the past for it's nuclear properties, should not be a problem unless you find out sanding it is i a good idea.
 
On washers: The washers I used are not BeO (as far as I know) - they are white, flexible plastic, yes, quite thick, but squish down when bolted up. I'd assume totally safe to handle. I believe the stuff is available in sheet form for you to cut to shape. Usual warnings though about not over tightening MOSFET/sink mounting bolts, to avoid distorting the device and losing a good flat contact with sink. Perhaps more so with these washers as I believe they are designed to give a little.

Insulation? Guess it doesn't matter too much at solid state voltages, but their thickness must be good for this. Can't remember the specs - check the web wite. Sorry - got their URL wrong above - should be http://www.bergquistcompany.com. (Got mine as samples from UK office)

Cheers

Pete

[Edited by peted on 02-24-2001 at 06:54 AM]
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.