I tried pulling D7 and D8. No change. As soon and the bias begins to rise it begins to oscillate. I have heard in the past that the MJL21193/4 have a tendency to oscillate and I experienced that with the Slewmaster. Valery suggested a fix for that but I had already removed them so I didn't get to try it. I hear all the time that these are good outputs so I would really like to find how to make them work. I have used them in small amps without issue.
Thanks, Terry
Thanks, Terry
Those devices probably work without issue in an EF2, but move up to a triple and added measures for stability need to be used. You could try just slugging the drivers and pre-drivers with some capacitance between the bases and collectors, maybe try 47pF on the pre and 200pF-ish on the drivers just to see if it settles down. That might not be the best or most elegant approach but might help some.
If I remember correctly his suggestion was 47pf caps on the drivers. My buddy's Parasound amps use these output devices. 2 pairs I think per channel. Nice amps. He's been checking out the Slewmasters I've been playing with and is starting to get worried.
I'm still learning simulation tricks. I can finally get that nice flat 'floor' on the FFT graph. I played with the simulation of this one some time ago and was never really satisfied with it. I will have to fire it up again sometime soon and see what else catches the eye, keeping OS's suggestions in mind too.
Terry, you are using the lower ft MJL2119x devices on this one if I'm remembering correctly. Seems to me triples in general work better with faster devices.
As for protection, I tried the simulation with all protection related components removed and it wasn't any better. Sketchy at best. I'm not thinking the protection is likely at fault, but almost anything is worth investigating.
Funny (actually not) that the designer, Struth, never came back with ANY support on this even after my sending him a set of boards. The same faulty board appears to be for sale on his website too.
Yes, Kevin (Struth) seemed like a good guy. He was pretty good about answering emails. The last reply I got for him was shortly after he received those boards. He said he wanted to build in one section at a time and test each one as he went. Sounded like a good plan to me. I sent him an email to remind him that the VBE traces needed to be modified and he replied that he did that as soon as he received the boards. That was the last time I heard from him. I sent emails every couple of months asking for updates but never got another reply. I should probably scrap this but I have three boards populated with a lot of fairly expensive parts so I would like to some day find the solution. I have built quite a few Blameless style amps now and all have been very satisfying except for this one. Funny thing, the whole purpose of this amp was a request by the OP for an amp to be used in DJ work or something similar. There is extra protection designed for and everything. Seems crazy that an amp designed for heavy use won't work at all. Too bad, I hear he is a respected designer. I'm confident he would have been able to get this working if I had just made the effort to build and test it.
I'm still hopeful OS will be able to help me get to the bottom of it.
Blessings, Terry
You notice the input stage / VAS is so very similar to the wolverine.
(It should work with a HUGE margin for error)
What is different - (I should of mentioned it) , is the triple EF.
Look to the slew triple ...
-4.7R stoppers on each output base.
-4.7R + 47-100uF separate decoupling for the predriver/driver.
-and ... 33pF B-C shunts on the predrivers + no resistor between
VAS and predriver.
What you have is a local feedback between output and predriver.
This was discussed in this thread .... http://www.diyaudio.com/forums/solid-state/149324-self-type-3-ef-hybrid-triple-any-pointers.html
If you are going to work with EF3's - read it. (Cordell + other "guru's")
PS - I almost forgot about the "magic triple" - its in the thread -post 103 !!
Terry , if you "tame the triple" ,the IPS should work perfect ... I did not
run square waves OR clip it last night.
OS
It IS the OPS ...
"Heavy use" ... as a radio transmitter , maybe ??
If you look at the slew-harmon kardons or any other EF3 , it is not just
"E-B to E-B to E-B". They all use stoppers/decoupling/x pF shunts.
Try the 33- 47pf shunts B-C on the predrivers first. I have seen these
soldered right to the pins of OEM pre-drivers (did they have stability
issues??)
The "next level" is either to have just the predriver or both the
pre and driver decoupled from the outputs with a 4.7R to 10R + a 10-100uf
cap (like the slew).
I see this on "high end" EF3's (HK990 -accuphase) . This slows
the triple down - but we have faster devices today.
Now that I think about it .... some of the old triples that use MJ15022/23
+mje15032/33 + mpsa92 DO NOT have these "tricks". They are slow
3mhz Ft devices - don't need any "tricks"
It is possible your amp was designed for these 1980 devices.
It is also possible the designer does not have experience with 30+mhz Ft
output stage devices - they DO need the "tricks".
OS
I tried a couple of things in spice. I found if I changed R2 to 33K as well, the offset dropped back to 9mv. I will try that as well in the morning.
I'll report back.
Who knows, if I can get this working, someone may want to try it too. I still have an extra pair of boards.
Blessings, Terry
I'll report back.
Who knows, if I can get this working, someone may want to try it too. I still have an extra pair of boards.
Blessings, Terry
yes, the effect is more current drawn from the input ltp base, forcing a lower output offset..
i still have some things i do not like about this amp,
the cap C4 in series with the 1k feedback resistor is quite huge....a 47ufd in series with a 1k resistor corners at about 4hz, this is quite low already...
remember that a cap at power on is like a short circuit and takes quite a while to charge up, this is probably what you were observing in your amp...
can't find them in the scheme you posted above #715 so i can not comment...
You need to go to post 686 to see those or look in the spice file. They were parallel with R16 and C19 took Q8 collector to ground. Looking forward to trying the caps on the predrivers. I have a 320u in C4. Originally I had a 4700u there per Struth's instructions. I may reduce that further. I don't think the diodes there are hurting anything now. The original blameless had one there.
Thanks, Terry
EDIT: the asc file is in post 708
Thanks, Terry
EDIT: the asc file is in post 708
Last edited:
even 320 uf to me is still big, more like 32uf and you do not need those diodes anymore.....
the hazard that a big C4 brings is LF instability and slow recovery from overloads....we want an amp that is fast to recover from overloads...thus a smaller C4..
i can not see why removing C11 and R56 affect output offset, they are not passing dc bias current at all....
C13 is 820ufd and connects from positive rail to ground, again no connection to output offset voltage.....
unless in your board there are slightly conductive in between traces offsetting the virtual ground output node...make sure that your board is clean by swabbing it with cotton buds and liquid cleaners such as isoprophyl alcohol...
the hazard that a big C4 brings is LF instability and slow recovery from overloads....we want an amp that is fast to recover from overloads...thus a smaller C4..
i can not see why removing C11 and R56 affect output offset, they are not passing dc bias current at all....
C13 is 820ufd and connects from positive rail to ground, again no connection to output offset voltage.....
unless in your board there are slightly conductive in between traces offsetting the virtual ground output node...make sure that your board is clean by swabbing it with cotton buds and liquid cleaners such as isoprophyl alcohol...
Forgive me I stated the wrong caps. I meant to say C11 and C19. But what I discovered in spice is that what caused the offset was changing R16 to 33K without changing R2 to 33k. I am hoping to find that is also true in fact when I change it on the board. I will also try putting a small cap in C4.
Thanks, Terry
Thanks, Terry
R2 and R16 are required to be identical, R2 being connected to psu ground is also at ac ground, R16 is also connected to the virtual ground formed at the output of the amp, with no input signal it is at ac ground or zero volts...
what happened when you changed R2 to 33k is offset this balance in such a way that the output node is closer to zero volts than otherwise...this means that the imbalance is happening elsewhere.....
again, i can not see how C11 and C19 can affect output offsets, they are there for ac signal compensation, not for dc biasing of the amp...
it looks like this amp has a lot of bugs that Struth needs to address to make this amp viable...
what happened when you changed R2 to 33k is offset this balance in such a way that the output node is closer to zero volts than otherwise...this means that the imbalance is happening elsewhere.....
again, i can not see how C11 and C19 can affect output offsets, they are there for ac signal compensation, not for dc biasing of the amp...
it looks like this amp has a lot of bugs that Struth needs to address to make this amp viable...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.