Need help with Giesberts HEXFET AMP

[ilimzn]

Hi Zimo,

Notwithstanding all the suggestions made to date, partticularly those from Ilimzn, the circuit has a flaw!

R10, R14 need to be linked out for the compensation on the first stage load to work!

Cheers,
greg

How so, don't the RC's effectively short any differential current in the collectors of the LTPs, hence no AC component to the bases of the VAS? OTOH, I'm not myself today with the mother in law of all colds, am I missing something?

 

[Zimo]

I turned my amp on yesterday!

It works now (again)

I didn´t change those 2.2nF caps to 2.7nF original but however I took the risk.. So it seems really that the biggest problem here was those IRFxxxN pairs. I haven´t measured it yet with a scope, but now listened it about three hours. Sounds great

Bias current was easily increased to more than 500mA with those changed resistor values. So high value is not necessary so I left it to about 200mA. DC offset is about +-3mV per channel, so quite good indeed (It could be easily adjusted to -+1mV if you put there 10- or 20-turn trimmer.)

My DC protection circuits seems not to working yet.. I tested it with 1.5VDC but it didn´t took speakers off, but left one cuts speaker when it gets warmer.. (The layout is design of my so the problem can be there)

So there is few "bugs" left that I´ll have to repair

But anyway Thanks for anyone of you (specially to ilimzn), IRFP models were the real key to solve the problems.

I´ll tell you more info when I measure amp with scope and generator.

See you soon.

Simo

 

[amplifierguru]

Hi Ilimzn,

I think your cold has got hold!

..."How so, don't the RC's effectively short any differential current in the collectors of the LTPs, hence no AC component to the bases of the VAS? "...

Looking at the upper half, the output from the first stage is single ended not differential - the voltage across R9, not R9 and R10. So the compensation should be across R9 to the rail. Since there is no output from R10 to the next stage the easiest way of ensuring the comp (R3 series C3) is across R9 is to link out R10. The compensation as it stands would be fine for a differential output from the first stage but the 1.2K of R10 adds to R3 = 1K247 rendering it ineffective.

This is also why the cct is insensitive to C3 at 2n2 or 2n7 - or 1n or 10n!! It's not doing anything.

Hope that helps.

Cheers,
greg

 

[Zimo]

I got my DC protection circuits work!

The problem was that the chematic had changed places of two resistors. However now it cuts speakers of when +-0.8VDC is running in the output. And the temperature problem is also gone.

Thanks for your explanation amplifierguru

I can´t understand your "fix", since when you took R10 off the T2 collector is not connected to +35VDC anymore? and I think I can´t work without that... or can it?

Anyway I changed C3 and C4 to their original values. Since I can´t measure the difference anyway I have to just trust that it cuts now differentialpairs frequency at 6.5MHz.

Originally posted by amplifierguru

Looking at the upper half, the output from the first stage is single ended not differential

I found this explanation based on Elektors original release.

"RC networks from the R3/C3 and R4/C4 decrease the bandwidth of differential amplifiers and power amplifier in the 6.5MHZ. Resistors R11-12-15-16 function as local feedback in the differential amplifiers improving the linearity. The differential amplifiers are supplied with constant current from current sources T5 and T6."

 

[amplifierguru]

Hi Zimo,

I was not clear explaining! When I said -

..."R10, R14 need to be linked out for the compensation on the first stage load to work!"

I should have emphasized LINK! R10 and R14 need to be shorted. As it is it has < 6 dB effect so almost any C can be used there or pull it out. But it's presence is intended to form a pole at fp=1/(2*pi*R9*C3) and a zero at fz=1(2*pi*R3C3) but this zero is currently at fz=1/(2*pi*(R3+R10)C3 which is the same as fp because R3<<R10= R9. So C3 just parallels R9 and R10 at Fp with little effect.
The amp will not function if you pull out R10 and R14. LINK them out.

Hope that's clearer.

Cheers,
Greg

 

[Thijs666]

Zimo, I was browsing your website and saw you own a Yamaha CA-1010. I have a CA-1000 in my junkbox, you don't happen to have any schematics of the 1010?

 

[Zimo]

Glad to see that someone else have also great old Yammies

Yes I have schematics to CA-1010 and as far as I know CA-1000 has nearly the same interiors, so I quess they will fit to your one also.

I just got a scanner and will scan those as soon as I can. Just give me your email address and I will send those pictures to you

I´ll try to make a PDF to compress that a little if I can. (haven´t done before)

BTW what is the problem with your amp? Mine had a collapsed transistor which driver relays and It cut the speaker relay, when the transistor warmed. Not a bad problem but so hard to find (this seems to be a usual problem with 70´s Yamahas)

Cheers!

Simo

 

[padamiecki]

I got my DC protection circuits work!
[....]

Hej Zimo!
thank you for your posts,
and all of you that sent yours advices!
I burned ~15 pairs 540/9540 during messing with this circuit...
Could you send real-working schematic with measured voltages of your prototype?

 

[Rick NL]

Hi Guys,
This amps seems to be sensitive.
A few things.
Your data refer to 540/9540 and the PSupply on 2 x 30 VAC 2 x 43 VDC. This may be too much for these rather small devices. The original Hexfet used 2 x 25 VAC and the IGBT 2 x 30 VAC.

In the July-Aug 2002 Elektor a revision to get better stability from both versions was published.
Place 27 nF paralel to R31
Place a resistor of 20k between collector of T8 and mass
Place a resistor of 20k between collector of T9 and mass
Replace R20 by 1k8
Replace R17 and R18 by 390 Ohms
Replace R3 and R4 by 33 Ohms (same sort of effect as jumpering R10 and R14?)
According to Elektor this should improve the stability and the performance.

Hope you can use this information

Rick

 

[amplifierguru]

Hi Rick NL,

"Replace R3 and R4 by 33 Ohms (same sort of effect as jumpering R10 and R14?)
According to Elektor this should improve the stability and the performance."

C3,4 create a pole with R9,13 at 50KHz (with R10,14 linked out)
and a zero at 1.2MHz with the 47R for R3,4. A change to 33R will only take that zero to 1.6MHz. But the original pole cannot occur because R10,14 get in the way STILL. They need to be jumpered.

Cheers,
greg

 

[Zimo]

Hi everyone!

To: padamiecki

Whouh! You´ve burned maybe 5 more than I (and I thought it was top of the iceberg already). First are you sure your FETs are the original 9540/540 pairs without the "N" marking at the end. I haven´t tested with the originals (which are much better complementaries than those with "N").

I also advise you to buy IRFP9140/240 pairs!!! These are good ones like ilimzn have told in this thread before, if you have read the all three pages. I have not burned any of these ones! (yet)

There is many problems with the stability of this amp. But It seems that when you make some modifications you have a steady amp. I have been listening my amp after changing to IRFP pairs maybe 200 hunred hours now. Only thing that have being screwed up now is on the Power On Delay-board.. those four 10 ohm resistors doesn´t seem to last long.. Anyway DONT BUY ANY FET EXCEPT IRFP PAIRS ANYMORE TO THIS AMP!

All the changes what I have done now, are listed in this thread. Read it thrue. I have not done yet that mod that amplifierguru is telling.. And don´t know yet am I even doing.. since this seems too good now.

Afterall I will post here an upgraded version of that schematic and it will include all differences I have done to my amp!

To: Rick NL

Thanks a lot mate!

I haven´t heard anything that Elektor have released any further data after IGBT patch..

"Your data refer to 540/9540 and the PSupply on 2 x 30 VAC 2 x 43 VDC. This may be too much for these rather small devices. The original Hexfet used 2 x 25 VAC and the IGBT 2 x 30 VAC."

Well I have used 25VAC transformers and those gives me that 35VDC, so I have the right components in that section.

I have heard that IRFP models can handle also 43VDC (which is used in IGBT version). So one of my friends is going to build this version. If it can´t take it straight he is going to use two (or more) IRFP pairs in paraller.

I´m sorry that I promised to give a new information of stability of this amp with some test equipments, but I don´t have a good signal generator at this time.. neither a good dummyload or speaker with such performance. I will post some data here when I can use such a equipment

And last a free tip which I assume every of you guys know: USE A QUALITY MULTITURN TRIMMERS!! I made a mistake here, so I have to get some bourns 10 turn trimmers.. this amp is capable to +-1mV steady DC offset! I have now about 200mA bias current and haven´t had any problems with that.. not even with heating

See You Soon!

Simo

 

[Thijs666]

Glad to see that someone else have also great old Yammies

Yes I have schematics to CA-1010 and as far as I know CA-1000 has nearly the same interiors, so I quess they will fit to your one also.

I just got a scanner and will scan those as soon as I can. Just give me your email address and I will send those pictures to you

I´ll try to make a PDF to compress that a little if I can. (haven´t done before)

BTW what is the problem with your amp? Mine had a collapsed transistor which driver relays and It cut the speaker relay, when the transistor warmed. Not a bad problem but so hard to find (this seems to be a usual problem with 70´s Yamahas)

Cheers!

Simo

The problem with my amp occurred when I switched from class AB to class A operation too slow (don't remember exactly what happened, since that was about 12 years ago...) Back then I tried to find the fault, but my knowledge of electronics was (far) too little back then. I came across 2 strange white ceramic kind of "resistor", could be fuses or something, which appeared to be smoked. I went to a local electronics store for a replacement, but the owner didn't know what these things where (although he had been in the busines for more than 20 years) and suggested I just tried some resistors. I did, but they where fried instantly after power on. I gave up by then, since I didn't have a clue where to look. (I still don't know what these things are )

But I'm still very interested in the schematics (as I already emailed you) Maybe I can get my CA1000 to work once again

 

[Zimo]

Hi!

So here I am again, open amp on my table. This time it´s not opened because of bad reasons.

I promised to make a modified schematic and here it is:

Modified Schematic

Here is the original for comparing:

Original Schematic

Modifications have been marked with RED, so you´ll be able to see easily what have been changed.

Here are reasons for these mods:

F1, F2 = Slow 5A Fuse, this value have been changed, because now the current is bigger thrue the fuse, when starting the amp (10mF cap is charged). This is maybe too big value.. Slow 2A will do it maybe also.

Fuses have been taken out from sourceline to prevent oscillation. I didn´t add source resistors, because I can´t find good ones without inductive structure. Now you can also use different types of fuses and it is not so choosy to quality of the fuse, because 2 capasitors are after it and will filtrate railvoltage.

This is not the best way to do it, because now almost nothing is preventing outputpair from blowing if it´s going to oscillate.. 10mF caps are going to make a huge current if IRFP pairs are shorted.

T12 = IRFP9140 This change is made to stabilise more the amp. It can handle the load better and is "slower" that N models and IRF models. TO-247AC case can move heat much better than TO-220.
It is more brutal than IRF models and a good complement to IRFP240 model.

T13 = IRFP240 Same as above with T12

R26, R29 = 120ohm These are gate resistors to outputpair. I have increased their values from 15ohm to 120ohm. I don´t exactly know how this makes the amp more stable.. Ilimzn could maybe tell this better than me.. but I quess it makes difference to current of the gates.

R25, R28 = 22ohm These two resistors adjusts bias adjustment range. My amp was after this modification able to get biascurrent of 1A if wanted. But I let it be about 250mA.

Z1, Z2 = 6.8V 1W Zener. These two zeners are added here and original schematic doesn´t have these. Zeners are for protection of outputpairs and will take out voltagepeaks that are higher than 6.8V.

C15, C16 = 100nF 250V MKP polyester caps. These are added physically close to outputpairs and 10mF electrolytics. These should work like filters at high frequencies and prevent amp from oscillation.

So this is the situation right now. I´m not going to change anything before I can see if it oscillates still. The modification that Rick NL mentioned could be worth to test! Because it have been tested in Elektors lab (I quess).

Please inform me if you can see something odd or wrong in my text and pics. Hope that this will help those whose amps are not working yet..

Warning these pics are big ones!
Here is some pictures of condition of the amp at this moment.


Cheers

Simo

 

[Zimo]

I have to add to my last post something...

When you´re looking those pics of my amp, you´ll be able to see that there are two extra boards in the amp. Those are DC protection circuits. I´m not sure what is the model or designer of those.. Because I got schematics to these from my friend.

DC Protection Circuit

But anyway... Circuit uses LM741 op-amps, 3 in both of the boards.

These are quite cheap to build, since everyone has some 741, and resistors..

This model should cut speakers out when dc in output is + - 0.6V. However there was a flaw in that circuit and it didn´t work.. two resistors had changed places. I have changed the values now in the schematic.. Those with red text! Now it works.. but with these values my circuits works when voltage is + - 0.8V.. so you´ll have to change some values to get it work with 0.6V.

I have done also a PCB to this.. but I think it has so poor design I won´t post it here now. Maybe in the future.

I am also working on a circuit which is a add-on to this circuit. It will give you a red signal-led to your front panel which tells you when speaker relay is not connecting. It uses a simple TTL logic. It will be ready some day!

Simo

 

[amplifierguru]

Hi Zimo,

Your stability is possibly aided by the C of the IRFP's you're now using being reflected back into the Vas and millered down, across the output to rail - just as my suggested change to 1st stage load comp would have effected.

good to have reliable stability.

Cheers,
Greg

 

[slbender]

Hi!

So here I am again, open amp on my table. This time it´s not opened because of bad reasons.

I promised to make a modified schematic and here it is:

Modified Schematic

Here is the original for comparing:

Original Schematic

Modifications have been marked with RED, so you´ll be able to see easily what have been changed.

Here are reasons for these mods:

F1, F2 = Slow 5A Fuse, this value have been changed, because now the current is bigger thrue the fuse, when starting the amp (10mF cap is charged). This is maybe too big value.. Slow 2A will do it maybe also.

Fuses have been taken out from sourceline to prevent oscillation. I didn´t add source resistors, because I can´t find good ones without inductive structure. Now you can also use different types of fuses and it is not so choosy to quality of the fuse, because 2 capasitors are after it and will filtrate railvoltage.

This is not the best way to do it, because now almost nothing is preventing outputpair from blowing if it´s going to oscillate.. 10mF caps are going to make a huge current if IRFP pairs are shorted.

T12 = IRFP9140 This change is made to stabilise more the amp. It can handle the load better and is "slower" that N models and IRF models. TO-247AC case can move heat much better than TO-220.
It is more brutal than IRF models and a good complement to IRFP240 model.

T13 = IRFP240 Same as above with T12

R26, R29 = 120ohm These are gate resistors to outputpair. I have increased their values from 15ohm to 120ohm. I don´t exactly know how this makes the amp more stable.. Ilimzn could maybe tell this better than me.. but I quess it makes difference to current of the gates.

R25, R28 = 22ohm These two resistors adjusts bias adjustment range. My amp was after this modification able to get biascurrent of 1A if wanted. But I let it be about 250mA.

Z1, Z2 = 6.8V 1W Zener. These two zeners are added here and original schematic doesn´t have these. Zeners are for protection of outputpairs and will take out voltagepeaks that are higher than 6.8V.

C15, C16 = 100nF 250V MKP polyester caps. These are added physically close to outputpairs and 10mF electrolytics. These should work like filters at high frequencies and prevent amp from oscillation.

So this is the situation right now. I´m not going to change anything before I can see if it oscillates still. The modification that Rick NL mentioned could be worth to test! Because it have been tested in Elektors lab (I quess).

Please inform me if you can see something odd or wrong in my text and pics. Hope that this will help those whose amps are not working yet..

Warning these pics are big ones!
Here is some pictures of condition of the amp at this moment.


Cheers

Simo

Hi,

Just another set of eyes and ears on this topic... First off, it struck me that possibly the amps could have blown if one of the original 2.5 amp fuses was already open, when he turned it on. I can't recall any amp I'd worked on that siond or so, unless there was a blown item somewhere to begin with...

OK, maybe that was not the case with this amp and this circuit. Next, I found looking at those schematics, well they seemed very familiar. So I checked a few, an early 1970's NAD I'm currently fixing is not too similar, but it does have those BD139/140's in there, overall, if you omit the details, this circuit is basically the same as the 1980's Hafler DH-120 or DH-220 if you forgive that the latter has two pairs of Mosfets in the output.

Basically, bunch of bipolars on the input and then Mosfets at the output; compensation-wise the Hafler seems to be much more complicated, but truly the basic circuitry is very similar. OK, you have an 'led' where Hafler put three ordinary diodes in series, etc. and the hafler used slower Hitachi Mosfets in TO-3 cases... One thing that did strike me in the more modern schematic, those 15 ohm resistors as Gate Stoppers. Worthless!

Most Bipolar/Mosfet amps use values around 100 to 150 ohms as Gate Stopper resistors. In the DH-200, Hafler used 220 ohms on each Fet, and also small caps to slow down the 'faster' N channel Fets. In the DH-220, Hafler used a 470 ohm Gate Stopper and a 680 pF cap on the 2SK134 part and 220 ohm Gate Stoppers on the 2SJ49 parts.

Makes me think that with the more modern IRF and IRFP TO-220 and TO-247 plastic types, which are likely much faster that the mid-1980's Mosfets, so one should have the extra compensation components in there, likely they are to keep a circuit of this topology stable into the 10 MHz and 100 MHz range.

So, that's my thought, - this Hafler Clone with modern parts wasn't truly thought out into the 100 MHz.+ range, maybe the designer didn't have a scope that would go into that range ?

Anyway, unless you have a O'scope capable of seeing into the 100 MHz. range, I'd use R26, R29 at 220 ohms for safety's sake, and 390 pF between Gate and where they meet with R31 and another 47 pF from Gate to Ground on the faster Fet.

Lastly, seems that zeners on the Fet Gates tend to be in some amps, and not in others. The Hafler Amps generally have zerners limiting the drivers, not on the Gates of the Fets; where Zeners are used on the Gates, I seem to recall values about 12 to 18 Volts. One should check if the lower voltage used here is prematurely clipping the circuitry in the audio range.

And best holiday wishes - for an amp that remains stable.

- Steven

 

[saczek]

HI,

I have a small problem with one channel.

PCB is almost as original - i add protection circuit for loudspeakers on uPC1237, some Zener diodes, I put smaller caps C11, C12 (4700uF) and decouple it with 100nF, etc.

The problem is with offset. I can't get 0V. When I turned on first time I burned T7 (BD139). So I changed the transistor and when I want to set offset the range is min max= 3,29V-7,98V. Diodes D1 and D2 I get in pairs, so voltage on R17 and R18 are the same - 1,1V. Resistor values:

R9/R13 804R/799R

R22/R23 56R/56R

R11/R12/R15/R16 22R/22R/22R/22R

R17/R18 508R/509R

I have only one pair of resistors which values doesn't match (in compare to other channel, the working one) - R21. In the bad ch I have there 852R in the working one - 757R (original should be there 1k, but I don't know If it has such meaning for offset)

Please help me. I have no ideas where look for mistakes.
thx

 

[Samuel Jayaraj]

This thread http://www.diyaudio.com/forums/showthread.php?s=&threadid=89526 has more info on the published mods.

I have 6 of these boards, but haven't yet tried the changes sent thru PM by rtirion as I was busy making Quasi's N-channel amp.

I am thinking of converting the Elektor Hexfet amp into an N-channel version along with the suggested modifications. Then one could use devices such as IRFP 460, 450, 350 and 260. Any opinions on such a mod?

 

[Zimo]

Hi!

It´s been a while since last visit!

And my old links doesn´t seem to work anymore because the old server is gone...

So, i put the whole project material to a new location which can be found here:
Whole project material

Amp has worked fine until the right channel started to drop off. Problem is located to relay or transistor which drives the relay. I got some new transistors and have to chance someday.

Simo

 

[Zimo]

Update: Pictures of this project are now located on imgur.com

Photo Album - Imgur

My earlier hosts are not working anymore. I'll hope this one will last longer

e: original Elektor pdf http://www.megaupload.com/?d=NXCDX8UP