I must say the way the schematic is drawn up I am a bit confused, as my eyes have troubles following the different paths.
I am guessing the bjt pair that latches may be tr17 & tr18, which via r36 can shut down the ccs at the other rail. I'm guessing that capacitor c12 is the one that should discharge to allow the latch to release. But of course I'm likely wrong.
The way the cascode is done on the ltp, and on the vas seems unusual and I'm not sure how this works.
Yes you are correct. The latch is made up of TR17 and Tr18 and the capacitor C12 discharges via R32. The LTP cascode is biased by the reference D4. This voltage is a little low beacause at about 2.5V RMS input the hard clip inverts due to overload. I use 10V zeners here now instead. The VAS cascode is biased at about 1.3V below the negative rail due to the combined drops across R36 and R40. This works but does clip a little early on the +ve excursion. When doing updates on PCBs I now bias the base of Tr4 independently at 2 junctions below the +ve rail using a BAV99 bypassed by 1n0. This does require several changes to the HF compensation. I also a diode charge pump to the latch to give a properly defined delay until the latch operates which I set to about 1/2 second.
Ok, now I see one mistake on that schematic. There are 2 R36 resistors, one that you just mentioned with R40, and the other is the one I mentioned before, going to the ccs from the latch.
Would you have an updated schematic to share with all those changes on it?
I also noticed something about the vi limiter on that schematic that confuses me as well. The 2 limiter transistors' emitters are not hooked to the output bus, which I think is probably because the dot is missing there. If not the case, I'd be curious to know why.
And as I understand it, the limiter doesn't directly act to trigger the latch. Rather it does so indirectly because the latch "watches" a side effect from the limiter on the vas.
And that is where I'm curious to know how that works. Would that time constant be the charge of C12 via R37 and R38, from the tr4 emitter?
Then what is tr16's purpose? The second part of the vas?
Very unusual topo, which is what makes it more interesting...
I have updated the schematic and hopefully there will be less errors. The VI limit is conventional the dot is now present where you expected it should be. The lower R36 is in fact R34. During the action of the VI limiter, Tr7 conducts. This causes overcurrent in Tr16 charging C12 and then operating the latch. The time constant could be not well defined and when I rework these PCBs I add reroute R37 to the +ve rail, add an LED current limiter from TR16 base to Tr5 collector and add a BAT85 in series with R38 as well as changing most values.
The cascode VAS has a very high impedance and the Rout even after the triple emitter follower was too high (tens of Ohms) to provide voltage drive before NFB. The addition of the class A buffer (Tr16) brings the open loop Rout to about 1 Ohm. This buffer has a clearly audible effect on listening tests.
There are other updates which I have outlined in a text document tagged onto the schematic pdf.
I am reluctant to post the schematic for the PCB updates as I have not tried them on enough PCBs as yet to be confident that the PCB will be always stable. Also when it comes to cutting tracks and grafting in SMD parts etc I would prefer the PCB come back to me rather than risk DIYers blowing the PCB and speakers too!
One interesting thing I have seen on one these PCBs tested lately is a slight negative drift of the quiescent current as the power devices became hot, no thermal runaway!
The cold Iq in the +ve rail is best about 22mA after about 1 minute.
Fit low ESR C4 and change C2 to polypropylene. C1, C13 uncritical as Rin is so very high due to R1 bootstrapping.
The cascode VAS has a very high impedance and the Rout even after the triple emitter follower was too high (tens of Ohms) to provide voltage drive before NFB. The addition of the class A buffer (Tr16) brings the open loop Rout to about 1 Ohm. This buffer has a clearly audible effect on listening tests.
There are other updates which I have outlined in a text document tagged onto the schematic pdf.
I am reluctant to post the schematic for the PCB updates as I have not tried them on enough PCBs as yet to be confident that the PCB will be always stable. Also when it comes to cutting tracks and grafting in SMD parts etc I would prefer the PCB come back to me rather than risk DIYers blowing the PCB and speakers too!
One interesting thing I have seen on one these PCBs tested lately is a slight negative drift of the quiescent current as the power devices became hot, no thermal runaway!
The cold Iq in the +ve rail is best about 22mA after about 1 minute.
Fit low ESR C4 and change C2 to polypropylene. C1, C13 uncritical as Rin is so very high due to R1 bootstrapping.
I'd be curious to run a simulation of this. But that schematic has no part values and no part types. The referenced document names don't point anywhere to be able to find them.
Is this one meant for the 2N3055 for the outputs?
Where is there a document with the missing info? values, parts...
What do those pcbs look like? Are they still being sold now? for diyers?
Is this one meant for the 2N3055 for the outputs?
Where is there a document with the missing info? values, parts...
What do those pcbs look like? Are they still being sold now? for diyers?
Here's a pair of VII modules :
http://4.bp.blogspot.com/-iIIijbzCE8g/UeVAGB5GbsI/AAAAAAAABSg/UhlGWswffbM/s1600/CIMG5570+PA.jpg
or 2nd pic down here Build a power amplifier from pre built modules
and here's an earlier Iss 5 :
Crimson Elektrik 170/4 Amplifier Module. Photo by AndyGoloskof | Photobucket
Now regretting selling that last pair.
One channel of my Crimson 1704 has blown up !
Judging by the sooty flakes around the inside, it was "on fire" for more than a few seconds. The T2A mains fuse blew to save the house.
I was experimenting with another amplifier and had temporarily pulled the speaker cables on one channel and left them dangling !!!!!!
Later the indicator was OFF.
I don't have any 2n3773, not Toshiba, not even fakes.
I wonder if Onsemi MJ15003, or MJ21194, would make an acceptable equivalent?
I do have spare two pair of drivers 2sa1606/c4159 instead of the fitted 2sa968/c2238, in case these are damaged as well.
Any thoughts?
Last edited:
Was there no HF load?
Aren't they supposed to be 3055?
I have so many 3773s in stock, from that "era", I'll put them to good use some day.
Maybe you can find more old stock, like the ones I have.
I wouldn't think those would pose any problems, they're not even much faster, so compensation may not even need tweaking.
check out my post #20 and the schematic about
http://homepage.ntlworld.com/bepowell/schematics/VII.pdf
and post #3403 about
http://www.diyaudio.com/forums/soli...erview-negative-feedback-341.html#post3188517
What about C7 and R29 (Boucherot/Zobel network) ?
Very suspect are additional follow for me:
1) input cap C1/C13 (why not an MKP and instead this two electrolytics)?
2) operation current for T17 is not independend from the hfe of TR18; why ?
3) what kind of speaker DC protect steps are in use (not to see in the schema)?
Also very important is to know the frequency response far away of the audible region (between 100 KHz and 10MHz, without the low pass cap at the input, C2)
Only a flat roll off (i. e. without a peak inside of the mentioned area) at all usual loads so as without external load saves an operation without unwanted oscillation and thus without damaged tweeters.
This response for amplifiers I have not see from no manufacturer until now. OTOH - this is actually the most important information in case with a reliable way of working without unwanted oscillation.
P.S. I don't understand this term:
and left them dangling
Last edited:
The crimson 1704 has the Output Zobel built into the PCB layout.
No, I have bought three different pairs, at different times, of CE1704 and all of them had Toshiba 2n3773 fitted.
If your PCBs are earlier than type VII , it would not think it worth it to fix them. If they are type VII you can use MJ15003 (as fitted by Crimson at one point). If the 2N3773 are Toshiba a good alternative match might be 2SD424 by Toshiba which are available via Ebay. I have also grafted in TO3P devices such as 2SC3519 although you do need to test carefully for stability with faster devices. I have been modding VII PCBs as mentioned above (March) and they work extremely well.
The answers to your queries are:-
1)With the extremely high Rin due to the bootstrapped R1, the input capacitors are not critical, using two polarised electrolytics back to back reduces stocking requirements. By all means use any non-polar type instead.
2) Tr17 has no operating current! It is used to shut-down the amplifier when the output stage is overloaded.
3) When DC protection for the LS is required , use an external device. I am currently trying out a passive circuit which energises a relay and self-disconnects the load.
Capacitor C2 is not used to "hide" a peak beyond the band it is to prevent intermodulation with spurious HF signals.
What is the query referring to R29, C7?
Brian Powell
1)With the extremely high Rin due to the bootstrapped R1, the input capacitors are not critical, using two polarised electrolytics back to back reduces stocking requirements. By all means use any non-polar type instead.
2) Tr17 has no operating current! It is used to shut-down the amplifier when the output stage is overloaded.
3) When DC protection for the LS is required , use an external device. I am currently trying out a passive circuit which energises a relay and self-disconnects the load.
Capacitor C2 is not used to "hide" a peak beyond the band it is to prevent intermodulation with spurious HF signals.
What is the query referring to R29, C7?
Brian Powell
Hi
I can only share my experience with the commercial models - 620/630/640:
1. There should be no problem using MJ15003 / 15024 / 21194.
2. The 2sa1606/c4159 work pretty well in the driver position.
3. Make sure to replace the resistors on both Zobel pairs - one on-board and another one at the speaker terminals, as they usually blow up when something goes wrong due to oscillations.
Cheers, Yair
Hello Andrew
I assume the speaker wires shorted together? The shut-down usually stops catastrophic failure but it still really stresses the output devices; especially if the signal is at low frequency as peak chip temperatures are higher. Let me know if you need a repair/upgrade.
I wonder how and why the now increasingly popular active speakers get along with very simple power amps and yet seem to outperform much larger passive speakers , not in bass range, as the actives come as shelf speakers, but in the musically important range of 150 Hz to 4000 Hz. I got a defective Dynaudio active 2 way, and managed to drive it with the old Quad 303, using the filter limiter etc. section of the Dynaudio to drive the Quad. It is really superb. Could it be the triple cascade just loves simple reactive loads, better than nasty loads of passive crossovers?
triples
Hi. I missed your question the first time I visited about triples. I have always used complemetary triples in the Crimson biplolar output amps both npn-pnp-npn as well as npn-pnp-pnp. The crossover region is very narrow. The output stages need about 10 mA in the power transistors, more than this is of little advantage. Overbias eg 50mA does lead to gm doubling but the distortion is very smooth when in this bias area. The thermal stability is so good that we have used 0R1 output resistors for some years. Almost no thermal feed back is needed as the bias spreader tracks the input drivers. As the input driver only dissipates a few milliwatts there is only a need to feed back about 1 degree C of thermal feedback from the main heatsink. A small copper wire and glue achieves this quite easily. Many amplifiers I an currently running have Iq set to 11 mA at 20C when run up to 70C the bias holds at below 35 mA. These complementary triples need careful selection of devices for stability. I am working on quadruples at the moment.
Hi. I missed your question the first time I visited about triples. I have always used complemetary triples in the Crimson biplolar output amps both npn-pnp-npn as well as npn-pnp-pnp. The crossover region is very narrow. The output stages need about 10 mA in the power transistors, more than this is of little advantage. Overbias eg 50mA does lead to gm doubling but the distortion is very smooth when in this bias area. The thermal stability is so good that we have used 0R1 output resistors for some years. Almost no thermal feed back is needed as the bias spreader tracks the input drivers. As the input driver only dissipates a few milliwatts there is only a need to feed back about 1 degree C of thermal feedback from the main heatsink. A small copper wire and glue achieves this quite easily. Many amplifiers I an currently running have Iq set to 11 mA at 20C when run up to 70C the bias holds at below 35 mA. These complementary triples need careful selection of devices for stability. I am working on quadruples at the moment.
I am looking for an amplifier with low idle current in the output, but with same sound quality as usual at hard running amp. Under "class B standart circuit" by the URL
Class-B output stage with feedforward.
you will find in the second display image the distortion result of 10mA and in the third image those of 60 mA quiscent current through the output stage.
Thus you are right with your estimate.
Very interesting seems to be the result in the last image (feedforward EC).
An example for realizing you will found here:
MJR9 (MOSFET's)
and here
http://www.angelfire.com/ab3/mjramp/newamp2.html (BjT's)
Do you have any experiences with this kind of aditional distortion reduce and is this possible with help of an upgrade kit for your power amp devices ?
Thank you for advices.
Last edited: