| steevo |
I have been testing my stochino amp that I built and never finished many years until recently when it was first published, and so far have destroyed a IRF9640 and a IRF640 mosfet transistor. I have been testing it using square and sine waves up 100khz the specs say the -3db power point is 120khz so it shouldnt have any problem with this. I was only drawing about 2.1 amps from recomended power supply and driving a 4.7ohm resistor load. The amp should deliver 120 watts into 8 ohms and probably more into 4.7ohms So there was plenty of capacity left. I was Luckily the mosfets did not destroy any of the eariler stages when they failed but its still annoying. The mosfets were faily well matched. Maybe I had a poor batch or that these are just poor and ill suited for this application. Has anyone had any similar experiences.
Steevo |
|
|
| HBarske |
I don't think building a 120W/8R amplifier with a single pair of output devices is a good idea. And IRF640/IRFP9640 aren't an ideal choice, as TO220 devices can't dissipate very much heat.
You won't come even close to the 150 Watts the datasheet promises, no matter how big your heatsink is. |
|
|
| jacco vermeulen |
The 1984 StoCChino* amplifier i recollect does state TO-220 devices, but has ~35Vdc rails.
The Stocchino 300V/uS design, printed in Electronics world in 1997, does 120W/8 but has 2 pairs of TO-220 irf640/irf9640
* CCh=k, Ch =g
(can't wait to read your lines on the DiS article, Holger. A shame the mag always arrives so late here) |
|
|
| steevo |
| The power amp circuit has 2 pairs of mosfet outputs, see the attached diagram. |
|
|
| smithy666 |
I successfully built this amp several years ago and have been using them without problems since. The only issue I had was when first setting the bias current I didn't let the amp warm up enough and then wound the bias current up too much, which then caused a catastrophic failure. You only ever do that once...
Second time around worked like a charm. I'm pretty sure I ran a sine wave through the amp right up to 120W output (I used the higher 58V driver rail voltage) without any problems (but not at 100kHz). The power supply is regulated 55-58V for the driver stage and unregulated 48V for the output stage.
I have a question - why were you testing at 100KHz with >100W output? If you wanted to test the bandwidth, why not just run a watt or 2 through? From what I've read of other designs, few would stand up to such a test either.
Good luck - well worth fixing, as these are great sounding amps. |
|
|
| steevo |
If the amp has been designed properly then it should full fill its specification as with any well designed and engineered piece of equipment. You dont buy a car that claims 0-60 in 7 secs only to find that it blows up if when you try it or instead of 7 secs you only get 8 seconds. I was simply testing it thoroughly cheking its performance.
Steevo |
|
|
| smithy666 |
Steevo
I understand what you are saying, but how long were you driving the amp at this power before they blew? John Atkinson from Stereophile tests all amps by driving them at 1/3rd of their rated power for an hour (within the audio bandwidth), which he describes as the worst case scenario for an amplifier. Most are too hot to touch and ones with thermal protection shutdown. So while amp specs say 100W into 8 ohm, you can't possibly drive a 20kHz square wave at full power for any extended period of time, unless you have huge heatsinking and/or fan forced cooling. How hot did the amp heatsinks get? Maybe they went into thermal overload? |
|
|
| janusz |
Hi,
IRFP640/9640 devices are not well suited for continuous high power application of that magnitude. Alex Soton reported problems with his Stochino amp running hot a few years ago on this forum. As I bought four boards of the amp as well we started to exchange ideas how to solve the problem.
Alex also intended to get rid of the NFB electros. It appeared that from the beginning both of us were somewhat suspicious about the use of IRFP640/9640 devices in this design. I suggested that IRFP240/9240 or SFH9240 devices would better dissipate heat.
Also the use of 2n5551/5401 input pairs created matching problems if the NFB electro was to be removed so we decided to use 2sa970bl/2sc2240bl instead. Alex managed to match their hfe within 1%, my batch allowed me to match them within 2-5%.
Anyway, after Alex IRFP640/9640s blew he increased gate resistors to 100ohm and used IRFP240/9240 pairs and all problems disappeared. The amp with 150mA bias was running cool, while before he could not go above 120mA.
As I had a few other more urgent things to do my boards still miss a few parts and amps are not yet running. Running the amp without NFB caps was less successful. Offset in one of his amps was almost acceptable at about 40-60mV but the other channel was above 250mV. I do not know if he has managed to fix that problem since then as he also had a few other things to do apart from experimenting with offset.
Most of his experience with Stochino amp is available on his site
www. soton.ac.uk
cheers, |
|
|
| smithy666 |
Hi Janusz
I saw on another post that you (or maybe is was someone else on the thread) was looking at a way of matching devices. Have you seen the Elektor circuit (Giesberts) for the power mosfet matcher? I am constructing this at the moment. It tests 2 x N channel and 2x P channel at the same time. Was published in Nov 93 issue.
Can you also tell me the link to Alex Soton's site?
Thanks |
|
|
| janusz |
Hi Smithy,
Alex'es site is
www.soton.ac.uk
It's the addres I gave in the previous post but without that space.
Yes, some time ago I was looking for a good circuit to match Mosfets. I got one used by Nelson Pass, one from ESP and some other one. I did not see the one from Elektor so I'd be grateful if you could send me a scan of this circuit with some description. It sounds interesting.
my direct email is:
janusz.pradzynski@dpi.wa.gov.au
I hoped to have my first two Stochino modules finished by the New Year but first I had to fix my Yamaha c-65 preamp I bought on Ebay last year. It took me some time to fix all problems and when assemblig somehow I shortcircuited two resistors which were soldered vertically. If they were soldered with theother ends down nothing would have happened but in this case two emitters were shortcircuited and smal part of the board damaged by burning transistor. Hope to fix it this weekend. That will be no problem if the input transistors are still OK.
cheers, |
|
|
| smithy666 |
Hi Janusz
I've sent through a pdf of the mosfet tester circuit. I should have mine finished in a couple fo weeks so I'll let you know how it goes.
I'm also finishing some new cases for my Stochinno amps - the old ones are in 19" rack mount, but I've built from scratch some custom aluminium boxes (don't do it - hundreds of hours of work), which I am going to get anodised in a funky colour. So I'll be moving my modules over to the new boxes soon.
Let me know how you get on with the amp modules. |
|
|
| steevo |
I only ran the amp for a few seconds at that level and frequency with which I stated and it was not at full power.
I am going to make sure the mosfets are screwed tight down and have good thermal contact with the heat sink just incase that was an issue. The others survived.
I would expect to run the amp for a few mins assuming the temperature allowed it which it does. I have a big 0.7 deg/watt heatsink for each channel. |
|
|
| Samuel Jayaraj |
Moderators,
You would do well to change the name of the thread to the correct spelling of the original author. This will help when people do a search on the web in general and in this forum specifically.
Well, I have built several Stochino boards and tested them successfully. I have had no issues so far. However, I only completed one two channel amplifier, running +-55 volts for the front end (regulated) and +-48 volts for the output stage (unregulated). The output devices were IRF9640 and 640. I was not aware of matching issues about 5-7 years ago and I do not know how close they were.
The offset I got consistently from every board was in the range of 33-35mV despite tight matching of all other components including passive devices.
The difference was that I used BAW62 diodes instead of the specified IN4448 ones and I also soldered the TL431 on the track side and inserted this into a 5mm hole in the heatsink filled with heatsink compound. I never notices any thermal drift or problems with Iq. Ofcourse, the PCBs were my own design (hand painted and etched). I also used a pair of series connected IN4148s in anti-parallel with the DC blocking cap in the feedback network. The cap itself was again a bipolar made up with two 330uF caps.
Since this amp has a very high GBW, PCB layout and parasitics will play a huge role in stability.
I have no idea of the current state of the finished amp since I have lost contact with the friend for who it was made.
Perhaps, in a week or two I should have another amp or two ready and I will post some pictures, now that I have a fairly good quality Digital Camera.
As for the sound, this is the best I have heard thus far and I guess only Nelson's F4 might better it. The latter is my next project and I am only looking for the input JFETs.
Hope this raises hopes for the ones who are trying out the fantastic amp. |
|
|
| janusz |
Hi Sam,
Glad to hear from you again. Looks that the boards of your design might be somewhat better than those sold by EW.
I have decided to run the amp from +/-57V so IRFP240/9240s were my first choice. With the NFB cap in and matched 2sa/2sc trnasistors Alex pushed the offset below 10mV. Only without the NFB cap it was high. Alex did not use 1n4448 diodes but bax16.
cheers, |
|
|
| jacco vermeulen |
| quote: | Originally posted by Samuel Jayaraj
change the name of the thread to the correct spelling of the original author. |
Sam,
you took the words right out of my mouth.
A bit embarassing imo that someone with so many audio related patents on his name and who has contributed so many times to Electronics World hasn't earned to have his name spelled right.
A link to both the high slewrate Stocchino amp schematic and the ones for the (regulated) powersupplies :
www.angelfire.com/sd/paulkemble/sound8c.html
Collectors item, there's also a schematic of one of the earliest Hitachi lateral MOSFET power amp designs on Paul Kemble's place : the output stage of the HA-7700 from 1979 => |
|
|
| smithy666 |
| quote: | Originally posted by steevo
I only ran the amp for a few seconds at that level and frequency with which I stated and it was not at full power.
I am going to make sure the mosfets are screwed tight down and have good thermal contact with the heat sink just incase that was an issue. The others survived.
I would expect to run the amp for a few mins assuming the temperature allowed it which it does. I have a big 0.7 deg/watt heatsink for each channel. |
Steevo
Did you have it working ok before you ran 100k through it? Was it working ok at 1kHz at full power? I ran my modules at full power at 1kHz with no problem. Sorry if I am insulting you by asking, but I assume you had succesfully set the bias current on this module already and particularly after warmup? |
|
|
| steevo |
| I did set the dias current correctly to 50ma each pair. I ran the amp at 1khz but only at low power. |
|
|
| CBS240 |
| Once I was playing with an amp I made with a 100KHz square wave and sines of such frequencies and I noticed the output transistors were cross-conducting and large currents were flowing due to EF AB bias at high frequencies. I placed a 100uF cap from one base to the other, and alleviated the problem. The drivers in darlington also required such a fix, but only 10uF. Problem is the output capacitance of the transistor requires current to turn the device off at higher speeds before the other turns on. The cap allowed the negative driver to turn off the positive output, and visa versa by supplying an AC current path accross the DC bias. This was with 20MHz BJT's but the same problem exists with mosfets, in spades. Certainly if I removed the cap, the outputs would have released smoke if sustained 200KHz full power output because of overbias. |
|
|
| steevo |
| If anyone is interested here is the final circuit I used, I had to reduce it 1000 X 1000 pixels so some details is lost. I have the complete power amp article scanned in as JPG images and both sides of the PCB scanned in high resolution. If anyone wishes I can email them the complete article circuit diagram and PCB images. |
|
|
| AndrewT |
Hi,
my rule for predicting maximum output power from an output stage into a reactive (speaker) load is max Pout<=total device Pd/6.
your 2pair of 640/9640 have about total capacity of 480W. This would indicate 80W into a severe speaker load.
+-48Vdc rails and 8ohm load should give about this level of output.
I use a second, but usually less demanding, rule the output voltage into half load RESISTANCE should be about -0.5db. i.e. 80Winto 8ohm becomes 140W to 150W into 4r0. (or 136W into 4r7 @ -0db).
Both these tests MUST be done with the heatsink at normal operating temperatures. This usually means testing for just a second or so to get the scope view and then on the next run get the unclipped voltage readings (upto 20kHz). If you want to blow up an amp then the best way is sustained power into low loads with the heatsink temperature rising way above design temperature.
BTW. 0.7C/W is a medium rated, but adequate sink for domestic listening, IF the output bias is kept low. A FET amp demands high bias and as a result demands a big sink. |
|
|
| steevo |
This my reasoning and logic behind why the Amp should be able to run at full power.
The maximum total power that can be dissipated by the mosfets is 480 Watts.
Assuming a Mosfet case temp Tc = 80 degrees then they are derated to 60 watts
Per device. (25 = ideal mosfet temp) Total power that can be dissipated is (125-55) x 4 = 240 watts.
Using ohms law the mosfets dissipate maximum power into 8ohms when the output
into the load is about 17.5 volts rms about half the supply voltage. So the maximum power
they have to dissipate is into 8 ohm with a constant sine wave is about 38 watts.
At full power of 100 watts into 8 ohm and output voltage of 28.3 volts RMS the output stage is only dissipating 23.6 watts.
So given my 0.7 degree/watts heatsink dissipating 40 watts the temp rise is 28 degrees.
28 + 25 = 53 deg and a temp difference between the heatsink and Mosfet case of
20 degrees is a worst case I think gives a Tcase = 73 degrees.
So the amplifier should be able drive full power all day long in my view.
The Mosfet will operate up to a maximum of 125 degrees realistically.
Even into 4 ohms the output is required to dissipate 76 watts at 17.6 volts rms.
derating each mosfet to 50 watts per device this gives a total of 200 watts
for the output stage. The mosfet case temp would be 100 degrees in this case.
I admit that driving a square is different and its effect are different on the mosfets
Than sine wave. Although I’m not sure what the effect of 20khz square wave has as opposed to a
20Khz SIN wave. |
|
|
| AndrewT |
Hi,
you're logic is flawed.
I just ran a 120W pair @ Tc<=73degC for 80W into 8ohms and it sits right on the SOAR limit for 60degree phase angle load.
The 4ohm load has to be reduced to 24degrees to fit the 73degC SOAR.
If I reduce the Tc to 50degC (the worst out of the set of 4) then 4ohm can go to 38degree phase angle (still not enough to drive a speaker.
But your 116W into 4r7 load @ 10degree phase angle is well inside the 50degC SOAR by a factor of 2.
I need to increase Tc (worst case) to 100degC to bring the SOAR down to the device peak dissipation.
And that I suspect is the problem. The devices are exceeding a catastrophic temperature limit during your high frequency testing, due to a combination of dissipation into the load and cross conduction that is independant of load.
Have you measured the operating rail voltages just before destruction? I could run the model at that PSU voltage and see just how hard you are stressing the devices. |
|
|
| steevo |
| The power supply voltage at failure was +-45volts and current was about 2.1 amps |
|
|
| AndrewT |
Hi,
using your +-45Vdc for loaded rails, the model shows 165W into 4r7 at SOAR limit for 80degC. Peak dissipation is shown as about 130Wpk/pr
The power input would be 189W and output is 165W, therefore only 24W is sent to the heatsink (I don't believe this, it's more usually about 60% of output power). This would indicate a sink temperature of just 45degC and a Tc of about 55degC.
There seems to be a small margin (80-55=25Cdeg) at normal frequencies. But the high frequencies probably took the devices over their peak capability, or the sink was hotter than the model predicts.
However, if one substitutes 60% of 165W for sink dissipation, the Tc temperature jumps to about 110degC. That more likely explains the blow up, particularly if there was cross conduction to increase device dissipation a little/lot further. |
|
|
| TomWaits |
| quote: | Originally posted by Samuel Jayaraj
Moderators,
You would do well to change the name of the thread to the correct spelling of the original author. This will help when people do a search on the web in general and in this forum specifically.
|
Attention moderator on duty, here is the gentlemen's name:
Giovanni Stochino
Cheerio,
Shawn. |
|
|
| DavidK |
Hi Steevo,
I see in another thread that this amp is compared to the Leach amp in terms of having a high slew rate. I have built the Leach on Prof. Leach's boards and am very happy with the performance but am interested in this amp too. I would be grateful if you could e-mail me all the info you have on it.
Thanks in advance
Regards
DavidK |
|
|
| steevo |
| ok what is your email |
|
|
| steevo |
Can anyone help,
I have been looking at the SOA graph and info for the irf9640 and irf960 and not quite sure how to inertpret the SOA data for the amplifier output stage. Using a worst case supply voltage of 100 volts as recomened by R. Slone ie (2*50volts) for inductvie loads, 100/8ohms = 12.5 amps peak for ech half cycle. divide by 2 for each transistor ouput stage half gives 12.5/2=6.25 so ech transistor has to handle 6.25 amps peak. Drawing a line acros horizontaly so that it intersect the 100 volt line show that it enters a region were the transistor can only withstand this current and voltage for 1ms. Have I interpreted it right. According to the graph VGS has to reduced to less than 20 volts for the transistor tro operate continously without damaging it. Is this correct.
Thanks |
|
|
| steevo |
| I forgot to attach the graph so gere it is |
|
|
| AndrewT |
Hi,
you should do a series of calculations to determine the load line for your loading.
Then de-rate the appropriate SOA for the temperature of the device cases.
Then compare for your operational conditions.
Look up Bensen's spreadsheet and find the reference in there that shows in detail the manual method of what I have quickly described.
If you can't find it, Email me. |
|
|
| jacco vermeulen |
| quote: | Originally posted by steevo
Using a worst case supply voltage of 100 volts. 100/8ohms = 12.5 amps peak for ech half cycle. |
Using both criteria at the same time is wrong.
First step would be to derate for the temperature rise. |
|
|
| steevo |
Going back to what I said about testing the Giovanni Stochino amp at full over a host of sin and square wave frequncies, I have a quad 306 and the service manual for the unit gives various test at frequencies and powers, one of the test is testing at full power 50 watss into 8 ohms at 100Hz, 2khz, 3khz, 10khz, 20khz and measuring the distorion.
Other tests include 70watts at 4ohms and a test at 25watts into 1ohm at 1khz. If the quad 306 could not perform as specified and failed when driven then it just would not sell and people would not buy it. Going back to my point about the Stochino amp the specification given in the article,
80volts peak to peak at 20khz giving distortion 0.0170%
Power bandwidth of 80Khz.
It should definitely be capable of delivering the performance quoted |
|
|
| smithy666 |
I seem to be one of the only people who has built this amp exactly in accordance with the design. I have been using a pair of these for over 8 years and never had a single problem with them. They have been driven hard. I also undertook extensive load testing when first built and as I recall drove them to clipping point with an 8 ohm load attached (which was approximately 120W). I also recall running at this level for minutes. Nothing went zap or bang and the heatsink (or transistors) weren't melting. The heatsinks I used were 0.39degC/W and they are never too hot to touch. My main speakers I use with these are Dynaudio Finales, which are not the most efficient in the world, so the amps run at a reasonable level.
It sounds to me like you have a problem with your transistors, thermal transfer to the heatsink or a construction error. |
|
|
| smithy666 |
Steevo
Why don't you e-mail Giovanni and see what he has to say about your problems? I would be interested in his response + if he has done any follow-up work/mods. |
|
|
| AndrewT |
| quote: | Originally posted by steevo
Going back to what I said about testing the Giovanni Stochino amp at full over a host of sin and square wave frequncies, I have a quad 306 and the service manual for the unit gives various test at frequencies and powers, one of the test is testing at full power 50 watss into 8 ohms at 100Hz, 2khz, 3khz, 10khz, 20khz and measuring the distorion.
Other tests include 70watts at 4ohms and a test at 25watts into 1ohm at 1khz. If the quad 306 could not perform as specified and failed when driven then it just would not sell and people would not buy it. Going back to my point about the Stochino amp the specification given in the article,
80volts peak to peak at 20khz giving distortion 0.0170%
Power bandwidth of 80Khz.
It should definitely be capable of delivering the performance quoted |
Hi Steevo,
what is glaringly missing from those power specifications is the duty cycle used during the tests carried out by the manufacturer.
Some of these maximum and medium power tests are often carried out at duty cycles around 10%. This is simply to keep the heatsink temperatures down near normal operating temperatures but still see what happens during these sweeps. An alternative may be to specify the maximum temperature of the heatsink and for the service technician to allow time for cooling between runs. He won't like that, it ads to his costs. It's just possible the Quad can supply those power outputs indefinately, but that would be very unusual for cost effective domestic amplifiers. |
|
|
| steevo |
I agree that amp will not be able to provide those figures for sustained periods thermally. But for a few mins certainly enough to obtain the measurments. The test i quoted were done using a SIN wave I think. I dont know what the duty cycle was for the square wave test.
I will find out. |
|
|
| steevo |
With refercne to smitthys coments, I am going toreplace all the mosfets and make sure they are screwed down tight and re-test the amplifer. Can you give me what yo think would be appropriate testing frequencies and powers/loads for the amplifier.
Thanks |
|
|
| AndrewT |
Hi,
first you need an analogue oscilloscope (20MHz to 50MHz) to see clipping on the output and to see any high frequency oscillation on the output.
Fast digital scopes are rather expensive.
8r0 and 4r7 dummy loads of at least 100W dissipation, preferably more.
Apply the full power signal for only ONE or upto a maximum of TWO seconds and arrange a flick switch to increase and reduce the input signal by 20db. This way you can see the input and output signals on the scope when the amp is just ticking over and go to +20db to adjust the input until the output starts to clip. If this adjustment takes too long flick back and let the amp cool. Then readjust.
Now reduce the input signal until you are 0.05db to 0.1db below clipping. flick back to low power.
Now prepare to take a voltage measurement at full power.
Do all this at 1kHz.
If you are interested in how the amp performs delivering full power at other frequencies, then repeat the whole process for 100Hz and 10kHz.
Change your load and repeat the setting up procedure. and then the measuring process again at your chosen audio frequencies.
If you want to do frequency/bandwidth testing, then do all your testing at the -20db below full power or even a little less (-26db below full power).
I think your 2pair amplifier is only suited to 8ohm reactive loads (on +-48Vdc supply rails) so the 4r7 test to full power must be a resistive load. Do not add any capacitors nor use long speaker cables. For the 8ohm testing you can use parallel cap for checking for instability, many would advise this as normal amplfier be-bugging. |
|
|
| dexter |
Dear Steevo,
if you don't mind me asking, I would like to have the xtochino amplifier files.
Would you please send it to dexter@tekom.net
many thanks in advance. |
|
|
| ejno |
Dear Steevo,
if you don't mind me asking, I would like to have the stochino amplifier files.
Would you please send it to ejno@vp.pl
many thanks in advanc |
|
|
| janusz |
Hi Sam,
The PCB of your design - was it one sided or two sided? Would you mind sharing with us your design? Free or for some charge. As you did not experience any problems (eg HF oscillation) it was really good design.
I know that Bora was redesigning Stochino's amp to work with bipolars. I do not know if he finished this task. I saw only a few experimental versions. Hes version was also going towards notable simplification. Quite interesting. As most speakers are not very sensitive I prefer more powerful version of Stochino using IRFP240/9240 mosfets. It might be also interesting to try it with more powerful lateral otput devices such as top Exicon.
cheers, |
|
|
| nikwal |
Might not be "ideal" pair anthough the name i similar..
I think if I remember correctly I used 9140 och 240, or 440 and 9240, dont remember now..
And regarding 640/9640, go throw them in the sea, I've never been able to get them to be able to take som serious abuse.. |
|
|
| jacco vermeulen |
Depending on the location, taking the rails higher than 45vdc might pose a risk for the 9140.
I had excellent results with IRFP9140/IRFP240 on Greggy Ball's SKA GB50D amp and JAM's mosfet A-40 version of the Pass classic.
Thanks to Ilimzn, and the web wonder.
Watch out not to buy IRFP9140N, flush 'm if you have.
A lot of vendors offer the 9140s but they'll send you the N version of the IRFP9140. (even wholesalers, as i've found out the hard way)
I made the mistake before i read Ilimzn's warning on the 9140N, but it did save me from wasting plenty time on trouble shooting.
Initially, i also used the 240/9240 combo on the SKA. On electrostats, the 9140 is clearly the better sparring partner for the 240. Read the posts of the Croatian expert.
(IRFP9140 prices vary a great deal, it pays off to shop around) |
|
|
| janusz |
Hi,
True, IRFP9140 is a good match for IRFP240, notably better than the 9240 at voltages below 100.
There is also not a great problem to find a few good N-P matches for use at lower voltages and powers. The problem is if one needs higher (120V plus) and power (over 100-150W per device). So the question is:
what are the best higher voltage/power N-P matches available on the market today?
Any success stories here going beyond 240/9240 and 640/9640 combinations?
cheers, |
|
|
| jacco vermeulen |
You need G-man Kanwar for that question.
I've been fond of fully balanced amps ever since i heard the Burmester 850s.
The 240/9140 duo is good for +500 watts/8, especially nice with crossed feedback as used in Pass products, taking distortion levels quadratically inversed down.
Funny enough, Burmester named his sustained class A input stage trick X-amp technique. |
|
|
| Workhorse |
| quote: | Originally posted by janusz
Hi,
So the question is:
what are the best higher voltage/power N-P matches available on the market today?
cheers, |
Hi Janusz,
The higher voltage complementary Pair is IRFP350/9350.....400V +14/-11A devices From IXYS.......
Kanwar |
|
|
| jacco vermeulen |
Class G, Kanwar.
Class H may be more appropriate but they haven't made a movie called H-men yet. :clown:
(G-Men, James Cagney, 1935) |
|
|
| janusz |
Thanks Kanwar,
Have you got maybe a datasheet for IRFP9350?
I couldn't find it on the net. IRFP350 is rather popular (IR, Intersil, Samsung etc). Has a high gate capacitance, which has to be delt with. I did not know that there is irfp9350.
Just in case - where one could buy these IXYS MOSFETS on the net?
Thanks again,
cheers, |
|
|
| Workhorse |
| quote: | Originally posted by janusz
Thanks Kanwar,
Have you got maybe a datasheet for IRFP9350?
I couldn't find it on the net. IRFP350 is rather popular (IR, Intersil, Samsung etc). Has a high gate capacitance, which has to be delt with. I did not know that there is irfp9350.
Just in case - where one could buy these IXYS MOSFETS on the net?
Thanks again,
cheers, |
Janusz, sorry i don't have the datasheet, because the IRFP9350 is discontinued and is now obsolete. but the pair is now replaced by IXTH11P50/IXTH22N50P complementary pairs... |
|
|
| janusz |
Thanks Kanwar and Smithy,
Smithy, thanks for the link to your site. I have four original EW boards. Two are alredy partly filled in with components. I was curious about Sam's design - especially if it was one sided board.
How is your heatsing performing?
I intend to use very similar 30cm heatsinks with flange from Altronics. The flange will be at the right angle to the board - its bottom parallel to the bottom of the case so the fins will be up to maximize heat transfer to the outside.
The case is from Jaycar and is almost finished with PS bridges and a bank of filtering caps - 6x10,000uF/80V Nichicon KG. With two amp boards, regulated PS board and protection board there is no room for anything else so all the transformers are in a separate box already operational. Problem that I have no time to finish the amp at the moment.
cheers, |
|
|
| smithy666 |
Janusz
I have been using the same heatsinks for years and they work well - never run too hot to touch. As I mentioned to you, I am swapping over cases at the moment - I used 2 3 unit rack mounts from Jaycar, but have custom built some aluminium cases which are at the anodisers now (bright red!). I am changing over my caps to 4 x Mundorf 33000uF per channel + using IXYS bridge. Years ago I designed a PCB for the regulated power supply, using LT1033/LT1085 regulators. I would have to look back to try to find the files (it was done in Protel) if you would like them.
I have 4 x unused Elna Cerafine 10000uF/100V caps + I will have 16 x used Elna Cerafine 10000uF/100V caps that I want to sell if you or anyone else (in Australia) are interested in. |
|
|
| janusz |
Hi Smithy,
Thanks, but I have already designed a board for the regulated PS. I did it manually. I also have one board meant for the Silicon Chip amp (SC design) but it is not the best of designs.
I think you should have no problems with selling Elna Cerafine caps. These are among the fastest. Thanks for the offer but I already have 12 10,000uF KG Nichicons so I do not need more big caps at the moment. I believe I should have used at least 4x22,000uF per PS for two boards rather than 6x10,000uF. Maybe I'll go for an upgrade in a few years.
I remember that originally I thought about getting Sikorel caps, which are apparently the fastest (TNT tests) or at least were a few years ago at the ime of that test. Unfortunately their price was quite high so I opted for much cheaper KG Nichicons, which like Elna and a few other, were not that much slower anyway.
I never used IXYS fast bridges. I thought about using them but again the price put me off. So I'm using two International Rectifier 35A bridges - one for the positive and one for the negative rail - instead per box. But I am curious about your experience with IXYS bridges. Is the difference audiable? If yes I may go for 4x27A bridges when the time for some upgrade comes.
I use fast diodes in the regulated PS and in PSes of all other bits of my equipment such as preamps and active crossovers. Must admit that my ears cannot tell the difference but I know it must be there. Use of the quality caps and proper filtering in general makes a difference to my ears as well as some ICs (but not all) but the difference is not huge - except for the extremes of course.
As Stochino's amp will be much faster than any amp I have ever built the use of fast rectifiers might be ear justified. I'm curious. I have also never heard of Mundorf caps. How do these compare with Elna Cerafine?
The only esoterics I use are Black Gate NP caps in parallel with polys in the NFB of the Stochino amps but to hear their impact I have to wait until I'll finish the amp and plug it in.
cheers, |
|
|
| smithy666 |
Hi Janusz
I've been using Hexfreds for the unreg and reg supply. I haven't tried the IXYS bridges yet, but they are much faster - will it be audible? Not sure, but I'll let you know when I get them finished. Mundorf make great caps. German made, very high quality but about 3-4x the cost of Elnas and from the specs are capable of delivering huge current. The physical size also works better for my new case (short and fat rather than tall and skinny!!), so I can mount the regulated ps board on top
I thought while I was going to the trouble of rebuilding the cases I would go the whole hogg and change the PS. I've got Tortech in Sydney winding a couple of custom encapsulated toroidal transformers - each one containg three separate dual secondaries to power the reg and unreg amp + protection circuit -to cut down on space used.
I agree on a tight fit - attached is a photo of my current config, all in one case. Don't remember how I managed to pull this off, but the new cases are much bigger, so should be less of a problem.
I also used Caddock resistors throughout and polystyrene/polyprop caps where possible and rather than testing for HV diodes, used higher voltage version instead. I think the electros on board are elna cerafine as well.
Smithy |
|
|
| smithy666 |
| another shot - closeup of the board |
|
|
| janusz |
Hi Smithy,
That is congestion!!!
Looks that Mundorf caps are even more expensive than Sikorel and your Stochino is a no compromise solution. I use Vishay low noise resistors in audio paths, Milne non-inducive resistors in Zoebel and sources and all caps other than electros are so called non-inductive polypropylenes.
One question about that higher voltage low reverse leakage diode. I could not find any high voltage diode of not much higher recovery time than those specified. So what diodes do you use there? Where one could get them?
And one final question about transformers. I got my two 625VA toroids from Harbuch. Unfortunatelly one is humming too much and should be replaced. It does not have to be encapsulated as it will be in another box - only quiet. Does Tortech do small jobs? It has to be 2x40V under load about 625VA. How much do your trafos cost? Who is the best person to contact there?
cheers, |
|
|
| Samuel Jayaraj |
Janusz,
Yes, my board design is single sided. I will try to send you a copy of the tracks and component layout. It is hand painted and etched and hence, may not look pretty.
I got almost all resistors custom made for the Stochino amp. I still have a couple of hundreds left of most values. Wherever possible I have used Silver Mica caps. |
|
|
| G.Kleinschmidt |
| quote: | Originally posted by janusz
And one final question about transformers. I got my two 625VA toroids from Harbuch. Unfortunatelly one is humming too much and should be replaced. |
G'day
Would that be an Antrim transformer? I am almost about to order a few of these from Harbuch, but I don't like the sound of humming.
Cheers,
Glen |
|
|
| smithy666 |
Janusz
Tortech will make 1-offs with no problem, or sell you 1 off the shelf. The one I have being made is 2x40V/400VA, 2x50V/100VA & 2x12V/20VA in the one encapsulated transformer. It is $230+GST. For a 2x40V/400VA alone it is $180+GST, so that should give you a guide. Would be cheaper for non-encapsulated. Just e-mail them with your specs for a quote - tortech@ozemail.com.au
I looked at getting one made by Plitron, but they wanted to charge a US$500 engineering fee just to spec it and come back with a price....guess they don't really sell 1 or 2 at a time.
I used 1N6638 in place of the selected 1N4448. The 1N6638 has VR=150V, trr=4.5ns and seems to be a great swap.
I also changed the input filters - I don't like 50uF of electros in the signal path - I made c1 & c2 2.2uF, R9 470R, R12 4.7k and C24 1nF, and dropped R39. I usually wouldn't modify anything, and apart from the large electros, can't think why I went to the effort. I haven't compared the two input filters in listening tests, but my Stochinos sound good. Maybe one of the gurus on this forum could let me know if I have made a bad mistake doing this.
Janusz - what caps did you use to bypass the electros in the feedback loop?
Smithy |
|
|
| janusz |
Hi Smithy,
Thanks for the info. 1n6638 looks really good on paper. I've checked my ordinary sources (Mouser, Farnell, DigyKey) and none of them has it. Where did you get it from?
As an input cap I use 10uF/100V polypropylene. It's big but just fits in. I'm not using R39, R9 will be 470 or 1k low noise Vishay (I haven't decided yet), C24 is 680pF silvered mica and R12=R20. These will be either 10k or 8.2k low noise Vishays. I have not decided yet but most likely I'll go for 10k.
D25 and D26 are doubled to reduce distortions and as I said before C22-C23 are replaced with a non-polar Black Gate 220uF/50V and it will be most likely bypassed by 2.2uF Solen or Bennic or ME polypropylene. I have a few of these. These are standard speaker crossover caps.
All the other polypropylenes are WIMA. Filtering electros are bypassed by XION "non-inductive" caps and each the PS rail will be additionally filtered by 680nF "non-inductive" cap in series with 1ohm/15W to the ground. That is supposed to minimize the effect of internal inductance of the big filtering electros. Never tried it before. Main electros will also be parallelled by some larger polypropylene caps: 22uF or 47uF. Small pF caps are 500V silvered mica but in some cases (47pF) I might use polypropylenes.
If the NP (or polarised electos) cap in the NFB is kept and connection with the preamp clean, it may be possible to remove the input cap.
I have also reduced R36 to 3.9 ohms as I have a few of these. Theoretically one should go down to 1ohm to minimize the ringing but I don't think my ears will hear the difference. The coil's inductance will also be reduced to about 2.2uH - certainly I will not go above 3uH. If the amps are to drive safe loads then the coil and R36 can be removed. Apparently the removal of both input and output filters makes audiable difference but I know for sure that removing the NFB cap makes a difference. The problem is to set the offset below 10mV but without a servo circuit it may be impossible.
cheers, |
|
|
| smithy666 |
Hi Janusz
Trying to remember where I bought them from (it was over 8 years ago...). I've done a bot of a search and it looks like it might (almost) be obsolete now. Some are still around, but the asking price is US$10-40 each!!! I can't believe they were that price when I bought them - from memory they were less than A$1. I'll have a look around here to see if I have any lying around as I would normally buy more than I needed. If I find any I'll send you some over.
Smithy |
|
|
| janusz |
Thanks, Smithy.
1n6638 diodes are obsolete. You might have bought them from Farnell as Farnell is a distributor for Microsemi.
One can buy them by thousands I believe at reasonable prices per item from some wholesalers. I found a number of them. But still it will be heaps of money for the lot.
cheers, |
|
|
| smithy666 |
Finally nearing completion on my refit. One amp is finished & assembled. I replaced the output transistors as I needed to change the way they mounted to fit the new heatsink. Interestingly, IR have upped the spec on the IRF640 devices - increased current and higher max Tj (now 175 deg, up from 150) - since I used them about 8 years ago. Anyway, matched the output devices, fired it up and set the bias current to 150mA. Seems to be pretty stable (+/-3mA). Also DC offset seems to be pretty low.
Pic of finished case below - I hand built these with the exception of anodising (the colour is actually cherry red) |
|
|
| smithy666 |
| Another pic - inside prior to fitting the amp pcb |
|
|
| janusz |
Hi Smithy,
Congratulations on the first module. Very crowded inside.
The middle board looks like a regulated PS, the right one could be a speaker protection board but that is guessing. None looks to me like the original PA board. The main board I presume should be close to the radiator but I do not see anything there.
So what is in the box? What are these boards starting from the left?
cheers, |
|
|
| smithy666 |
Hi Janusz
Not quite as crowded as my last one....I can actually work on this one without pulling everything apart.
I took the photo before putting the amp module in, then last night screwed on the top of the case without taking another photo - I will for the second amp, which will be finished in a couple of days.
The board at the back left is a soft start/remote switch power on (so I can use the push button on the fron panel). This is my own PCB. In front of that is the custom transformer I had wound - potted toroid 2x40V 400VA, 2x50V 100VA and 2x12V 20VA from TorTech. Just in front of the toriod is an IXYS bridge. Front right is 4 x Mundorf 33000uF/80V electros. In the middle rear is a regulator board for the +/-57V regulated rails. Again, one of my PCBs from several years ago - uses LT1085/1033 and HEXFRED rectifiers. To the right of this is a Borbely DC protection, turn on delay. The amp module bolts to the heatsink (cutout in back panel), with the heatsink being a 0.4 deg C/W Fischer unit.
The amp board still uses Caddocks throughout (standard and power), met polprop caps and blackgate caps. As mentioned, I replaced the IRF640/9640s and BD139s as the heatsink mounting method changed for this case.
I made the case from 3mm aluminium panels, 6mm aluminium front panel, with a frame of 10mm square rod and a sub floor of 1.6mm aluminium, which everything mounts on. I drilled and tapped all the holes and countersunk all the screws for a flush finish (except the four cap heads on the front panel). Then I sent all the external panels off to the anodisers....
Two years passed and hundreds of hours later I'm nearly there. Looks heaps better than my previous one.
I should add that the case colour does look really good (even though many will not like it), particularly in its final resting place. |
|
|
| janusz |
Smithy,
did you listen to this one or not yet? If yes, how does it sound comparing to the previous version?
What's the offset reading?
I haven't touch my modules for ages. The first box with two modules needs 20 odd hours of work to be completed so practically it could be done within two weekends but I always have something else to do like working on the new home. My tape deck also need rejuvenation. Sounds like a neverending story.
cheers, |
|
|
| smithy666 |
Janusz
Haven't had a chance to listen yet - like you it's hard enough finding time to complete the amps, let alone listening...
I quickly measured the offset last night and it was a lowish 6-7mV (Stochino measured 32mV on his). I did match all transistors when I built it years ago, but I don't remember what the offset was back then. I'm going to fire it up again tonight to take a few more measurements and do a final check before some listening tests, so I'll report back. |
|
|
| smithy666 |
| Measured offset again tonight and it is sitting around 6mV, which is good. Connected it up and away it went. Need to listen for a while to reveal any differences from the refit/upgrade, but it did sound great (again). I'm running them through Dave Ellis' 1801b and also Dynaudio Finales |
|
|
| ilimzn |
| quote: | Originally posted by janusz
Hi,
There is also not a great problem to find a few good N-P matches for use at lower voltages and powers. The problem is if one needs higher (120V plus) and power (over 100-150W per device). So the question is:
what are the best higher voltage/power N-P matches available on the market today?
cheers, |
Try IRFP9240 and IRFP340 if you can find the latter. IR seems to have obsoleted a number of their older IRFP series FETs, which is a pity, because they have not replaced all of them by the newer N versions, or the N versions are not good complements. There are droves of N ch ones available still because these were used in far greater numbers than the P ch counterparts. Currently IXYS seems to offer the best selections to chose complementaries from. |
|
|
| jacco vermeulen |
| quote: | Originally posted by smithy666
the colour is actually cherry red |
Lovely case, i've added the image to my collection, Smithy.
Very pretty on the inside too.
Who dare say Aussies lack fancy taste, mate ? :clown: |
|
|
| ilimzn |
| quote: | | Originally posted by steevo I have been testing my stochino amp... using square and sine waves up 100khz |
I wonder if anyone has noticed the part I have emphasized above. high power testing with HF square waves is asking for serious trouble unless you take measures in advance.
- If there is cross-conduction, a square wave test at HF can destroy an amp in seconds. Years ago I helped a friend make a Stochino amp, but because we could not get TL431's here, I redesigned the bias network using a Vgs follower. In order to compensate for the poorer impedance of the Vgs follower t HF I bypassed it with a non-inductive foil cap (0.47uF IIRC), to prevent cross conduction because of base current rectification in the driver stage.
- If there is a Zobel network at the amp output, your amp may be working in a MUCH lower load than expected, and in fact very quickly destroy t he Zobel, or itself. In fact, once the Zobel is destroyed, it may oscillate destroying itself by cross-conduction, and it will be difficult to figure out what the real cause of the problem was.
- square waves which are not limited in slew rate, may have a much higher harmonic content than expected, even taking into account input RC filters on amps. This is because driving an amp from a generator means it is driven from a very low source impedance, making the input filter have the maximum possible corner frequency, and given the filter is first order, you may run into NFB loop instability problems. The instability itself may not be terrible - for instance, just excessive ringing on the output. But, this ringing may contain frequencies that are far above the ones in theinput signal, hence you rin again into cross-conduction and 'low load by zobel' problems. |
|
|
| smithy666 |
Finally finished the second amp refit. Bias again set to 150mA. DC offset was a bit higher than the other channel, around 23mV (the other channel was 6mV). Anyway, all installed and sounding great. A few pictures follow.
The first photo shows both amps installed in their final resting place. Gryphon Tabu CD player sitting on top. |
|
|
| smithy666 |
| Photo of module mounted to bracket prior to mounting to heatsink. In changing the case/heatsink I has to replace all the IRF640/9640 and BD139s. |
|
|
| smithy666 |
| Photo showing module mounted and wired up ready for bias setting. |
|
|
| smithy666 |
Photo of rear of amp.
Almost 2 years of rebuild complete :) Now for the next project :D |
|
|
|
