There must be something of that nature, otherwise I don't see what they're for.
Some RCA, some Motorola, some Sescosem, some Thomson, and some ST.
I have several 3055H from RCA, and I wonder if those are the 100V or not.
Some of the Motorola 3055s are S, and I'm not sure what those S are really.
I couldn't find my MJ2955, but I think those are from ST. I have them somewhere, just couldn't get my hands on them now. I have all those things in an excel sheet as a stock database, so I know pretty much what I have in stock, so I know I should have at least about 40 of those MJ2955s somewhere.
What was that about?
Well that would be an entirely different project and I'm not really looking into those right now.
Actually I'm not doing this just for myself. There are many diyers out there who have such old stock, plus many who live in places on this blue planet where they can buy them cheap.
So I'm also thinking about sharing this with others.
Not sure which they are, but I do have many from all those suppliers.
I didn't come across this before. But if I was testing, I wouldn't go beyond something like 30k anyway. There's no point.
Besides, I'm not equipped to do such testing. Trying to make a prototype would be a bit futile for me right now, as I couldn't really troubleshoot it.
To power a new amp after a build, I would have to rig something.
But one thing I will definitely want to do before building, is testing parts and matching them, so at least I will have small test stuff to do so.
I already have small test devices for capacitors, resistors and basic tests on transistors. Nothing too advanced, but should help matching.
One of my projects in line is to make a full blown tester/matcher similar to locky-z's curver tracer. But I wouldn't make that for windows only like he did.
Ouch! I haven't played with those, so I didn't know such details. Important details. Too bad. But still, they're still manufactured, I wonder who uses them.
What did you change?
No, but I'd like to see that. Very interested.
Cool. Do you have something to share?
I have made a sim of my own using the bryston overall topo, with some small changes, and for 3055/2955. Works great, in sims. Low thd, nice performance even with those slow old parts, and those sims worked with the same models as I tried out in this leach superamp sim...
I was going to post a few photos of some of my stash, but they all came out blurry, so we can't even read what's on the parts...
I wrote: "but note that the constant power slope goes to SOA limited at 50V"
That should have been "but note that the constant power slope goes to SOA limited at 25V
My simulations are all posted on here, just list threads that I've started.
We tried several solutions for the reverse Vbe issue in the Tiger thread but none were
good enough because the Tiger feedback is very low Z. The Tiger .01 and Bryston use
a triple with a much higher Z feedback path and then the diode fix works at least in sim.
I did not sim the Tiger .01 but the Bryston is very similar.
That should have been "but note that the constant power slope goes to SOA limited at 25V
My simulations are all posted on here, just list threads that I've started.
We tried several solutions for the reverse Vbe issue in the Tiger thread but none were
good enough because the Tiger feedback is very low Z. The Tiger .01 and Bryston use
a triple with a much higher Z feedback path and then the diode fix works at least in sim.
I did not sim the Tiger .01 but the Bryston is very similar.
Last edited:
I was thinking about building a curve tracer, you might want to look at the Analog Discovery
that provides a 20MHz scope (rated at 1 MHz), waveform gen, and Logic Analyzer for about
$269, I think you could test semis by using it to control a power supply or just boost the
output voltage with a DC coupled power amp:
http://store.digilentinc.com/analog...ope-logic-analyzer-and-variable-power-supply/
It has a 5V low current power supply that is also programmable in the most recent version
but not the previous. It will take some software to do it obviously.
that provides a 20MHz scope (rated at 1 MHz), waveform gen, and Logic Analyzer for about
$269, I think you could test semis by using it to control a power supply or just boost the
output voltage with a DC coupled power amp:
http://store.digilentinc.com/analog...ope-logic-analyzer-and-variable-power-supply/
It has a 5V low current power supply that is also programmable in the most recent version
but not the previous. It will take some software to do it obviously.
Last edited:
I didn't really make any major mods to the Tigersaurus. I subbed MJE243/253 for the
40409/40410. Upped some of the resistors to higher power, that's about it.
40409/40410. Upped some of the resistors to higher power, that's about it.
I tried deleting the series outputs and drivers and, as you said, it simulated fine. Went back
to series and after trying quite a few simple mods came to the conclusion that using the
output to drive the series transistors was adding more phase shift.
Decided to use the first EF directly driven by the VAS, as you can see in the attached sim.
This alone helped but was not a complete solution, while doing the edits I noticed something
else that didn't look right. The drivers were also connected in series, the main driver's
collector needs to go to a low Z half volt point so I connected it to the power device emitters.
This looks like how you normally connect an EF. I went back to the original file and tried
just this mod, it did not work, it needs both. It can be fine tuned in many ways
I patched it really fast so it is a bit sloppy but try it, it seems to work fine.
I called the file new because it is so different.
to series and after trying quite a few simple mods came to the conclusion that using the
output to drive the series transistors was adding more phase shift.
Decided to use the first EF directly driven by the VAS, as you can see in the attached sim.
This alone helped but was not a complete solution, while doing the edits I noticed something
else that didn't look right. The drivers were also connected in series, the main driver's
collector needs to go to a low Z half volt point so I connected it to the power device emitters.
This looks like how you normally connect an EF. I went back to the original file and tried
just this mod, it did not work, it needs both. It can be fine tuned in many ways
I patched it really fast so it is a bit sloppy but try it, it seems to work fine.
I called the file new because it is so different.
What I have in mind for device testing and matching, it more geared to do just that only, and nothing else really. So it would be something specifically made for the purpose, and do that well.
I would want the test and matching results to be memorized and saved, and the measurements taken on parts could be compared by software as well as visually on screen.
The thing is, none of the available systems that I know of can really do all that, and they don't really handle power devices.
I would want to properly test power devices as much as possible in their usable area, and have the ability to test their second breakdown limit (non destructively of course).
One other important thing to test for nowadays is the counterfeit. Those should fairly easily be detectable, as they're really inferior to the genuine, so the differences should be detected and those could be identified.
To me, a good testing rig should just plug into a computer via usb, and just about everything should be software driven.
The software should allow saving the measurement data for easy mass comparison.
Batches of parts could be sorted and best matches identified, small signal all the way to power devices.
So perhaps turning those simple series arrangements into real cascodes might work better.
I can see it works, much better than before, however did you take a look a thd?
The thd is quite high, and all along I've been thinking this high thd is due to the fact that the amp's feedback is helping correct a lot of if, but not enough to get a better thd result.
It does look like it's fairly stable as it is now, but to me the high thd hints at something going on, I suspect mostly in the output stages, that are too much for the feedback to correct enough for better thd result.
If you look at the signal in some places along the path, you can see it highly distorted and in some places there are lots of spikes, which get mostly eliminated by the global feedback's effect, but obviously some of those must be remaining to contribute to the high thd.
Ok, it does work, but what could we do to get better thd levels?
The thd rises fairly fast and gets high as we approach max power level, and I got well over 1% even before getting to the onset of clipping.
I see the clipping behavior is a whole lot better though.
Plus I noticed the clipping happens a little higher, making better use of the voltage swing capability from those rails.
I tried applying a few of those things to the other version with 4 sets of devices that I was testing for protections before. I converted the stacked vas device into a real cascode, hawksford style, and applied a few of your changes to the outputs, like removing the extra emitter res from the stacked slave devices and unlinking them to make each stack independant.
All those things allowed me to stabilize it enough so it finally worked even with the 3055/2955 models, plus I managed to tweak the compensation for a very comfortable margin, all while getting rid of most belts and braces caps that I had sprinkled all over the output stages before.
So that also worked out well, and I actually get a much lower thd result.
However the cascoding on the vas did impose extra overhead, so the clipping happens much earlier, and on the vas, not the outputs.
The clipping shows a foldback effect though, and I thought there was some sticking going on, and I aded a couple of bav21 clamps on the vas, which didn't change thing very much, even though thd went down a bit. Go figure!
If there is a good way to unstack the pre-drivers, and perhaps the drivers as well, that might be good too.
I have also noticed one other thing of concern, in the 4 sets version I tested, which happens only on the side of the mj2955s, where they're not equally sharing the voltage swing. The 3055s are sharing it almost equally, with the active ones being exposed to about 70V while the rest goes to the slaves. That's odd. Why are the 3055s properly equalizing their voltages and not the 2955s?
It's different alright. I actually am having troubles understanding this new topo.
It's quite different. The drivers are no longer stacked, and the predrivers remind me of the bryston method of the quadcomp.
But one thing puzzles me. Which are the slaves in the output stacks now?
The nested inner feedback loop also seems odd now, with the second set of predrivers and whatever they drive being outside of that...
I see the mj2955s aren't exhibiting the behavior I observed in that other 4sets sim, which is they're actually sharing the voltage swing mostly equally as their 3055s counterparts.
The high thd must be related to some tiny oscillations somewhere perhaps. And the feedback isn't enough to wipe those artifacts out.
And they very nicely wrote some softaware to do curve traces... check out the digilent scope forum. So all you need is the digilent box, an external beefy psu and a small ohm power resistor. Then you've got yourself a precision curve tracer. Plus the software their Engineers wrote is great - it's stores the readings on an excel in your desktop automatically!
The only significant problem is fortunately easy to solve - the control pin ( gate) of the DUT can't be driven beyond 5V unless you stick a lil 2x gain single rail opamp on their wavegen pin. Actually trivial stuff and I'm not just saying that.
Thanks! Great to know, there is so much on their web site it would have taken a while to find. I am borrowing one currently but that is probably enough to buy one.
The center power devices are the main ones.
I got to where it is now based on something I thought years ago when I saw the first stacked
output and that is, it sort of looks like it is chasing its tail. It really is not since
the collectors of the main outputs are not inputs to the devices but it just looks strange.
There is a 50/50 voltage divider on the main output in the original to produce the input to
the series devices.
The output comes from the EF which is unity gain so the output is just a copy of the VAS
output. I made a new copy of the VAS with the predriver with the 50/50 divider on the
emitter and that now drives the series devices. To me, it no longer looks like it is
chasing its tail since the paths are in parallel.
I would check the biasing (all) and open loop gain at 20 KHz again to see how much
gain there is for distortion reduction. Perhaps try BD139/140 in place of the MJE350/
340 especially in the VAS. I like everything in the output section to be no less than
500 mA rated devices.
All the devices can be replaced with modern types (subject to prototype test) for better
performance.
Thinking out loud here first thing I'd do to reduce distortion is to try Cordell's MJ21193/4
in the output. If that fixes it we can try to figure out what to do about the other models.
Another comp is to see what the model without the series connection does for distortion,
those old devices have beta droop, low Ft, etc.
The series devices are just there to divide the voltage, they don't need distortion
reduction, even the output bias for them is set by the main devices.
I think it is better to have all the predrivers take the full voltage as long as the parts can
take it and you watch out for power dissipated. Most of those driver/pre drivers that I
used can. I connected the new series drivers' collectors to the output, they could go to
the opposite rail, or the series half way point for more Vce if that helps distortion at all
Pdiss goes up obviously.
Last edited:
Attached is an .asc where I removed the series devices in case you want to try if for
distortion. It is a completely normal triple EF but with the old devices. It uses the BD
parts and would need the MJEs as a comp to the New file, but try it to also check the
BD's.
distortion. It is a completely normal triple EF but with the old devices. It uses the BD
parts and would need the MJEs as a comp to the New file, but try it to also check the
BD's.
Attached is a Quasi comp Leach Super amp that might interest you for those 2N3773s.
I'd probably add more pairs depending if you want to rate it for 4 or 2 ohms. I didn't
sub a 2N3773 model as we would have to find a good one but a Cordell MJ should work
fine just to prove it out.
I'd probably add more pairs depending if you want to rate it for 4 or 2 ohms. I didn't
sub a 2N3773 model as we would have to find a good one but a Cordell MJ should work
fine just to prove it out.
Last edited:
Please post the New version uncommented to do the distortion run, there is a lot going
on in your sim and I want to do it the same way. I'll take a look at it.
A few things to keep in mind, the Leach as I'm sure you know followed the concepts for
low TIM which are to minimize transient overload in the diff pair (heavy degen) and wide
open loop gain (minimal Cdom heavy VAS degen). You can see the heavy degen on both
the diff pair and VAS. In contrast Self uses zero degen on the VAS.
My gut feeling is that the MJE340/350 have too much internal capacitance for the VAS and
that one/both of the output models is not right.
on in your sim and I want to do it the same way. I'll take a look at it.
A few things to keep in mind, the Leach as I'm sure you know followed the concepts for
low TIM which are to minimize transient overload in the diff pair (heavy degen) and wide
open loop gain (minimal Cdom heavy VAS degen). You can see the heavy degen on both
the diff pair and VAS. In contrast Self uses zero degen on the VAS.
My gut feeling is that the MJE340/350 have too much internal capacitance for the VAS and
that one/both of the output models is not right.
I'm fairly certain that I used this 2N3055 model from Duncan's web page:
BJTs
years ago and it worked fairly well, not sure where he got it. Does it look familiar?
*2N3055
*Si 115W 70V 15A 20kHz pkg:TO-3 3,2,1
.MODEL Q2N3055 NPN(IS=4.66E-12 BF=360 VAF=100 IKF=0.25 ISE=3.339E-11
+ BR=2 ISC=5E-9 RB=3 IRB=0.001 RBM=0.4 RC=0.04 CJE=5.802E-10 VJE=1.2
+ MJE=0.45 TF=8E-8 XTF=1 ITF=3 PTF=120 CJC=2.121E-10 MJC=0.4 TR=2.55E-6
+ XTB=1 )
Once you find a reasonable 2N3055 model you can just change the NPN to PNP and have a
comp to work with. Sure it is a perfect match, but better than a defective model.
Just noticed that he has a PNP also, I have not checked it in any way:
*2N2955 [DM addition]
*Si 115W 70V 15A 20kHz pkg:TO-3 3,2,1
.MODEL Q2N2955 PNP(IS=4.66E-12 BF=360 VAF=100 IKF=0.25 ISE=3.339E-11
+ BR=2 ISC=5E-9 RB=3 IRB=0.001 RBM=0.4 RC=0.04 CJE=5.802E-10 VJE=1.2
+ MJE=0.45 TF=8E-8 XTF=1 ITF=3 PTF=120 CJC=2.121E-10 MJC=0.4 TR=2.55E-6
+ XTB=1 )
BJTs
years ago and it worked fairly well, not sure where he got it. Does it look familiar?
*2N3055
*Si 115W 70V 15A 20kHz pkg:TO-3 3,2,1
.MODEL Q2N3055 NPN(IS=4.66E-12 BF=360 VAF=100 IKF=0.25 ISE=3.339E-11
+ BR=2 ISC=5E-9 RB=3 IRB=0.001 RBM=0.4 RC=0.04 CJE=5.802E-10 VJE=1.2
+ MJE=0.45 TF=8E-8 XTF=1 ITF=3 PTF=120 CJC=2.121E-10 MJC=0.4 TR=2.55E-6
+ XTB=1 )
Once you find a reasonable 2N3055 model you can just change the NPN to PNP and have a
comp to work with. Sure it is a perfect match, but better than a defective model.
Just noticed that he has a PNP also, I have not checked it in any way:
*2N2955 [DM addition]
*Si 115W 70V 15A 20kHz pkg:TO-3 3,2,1
.MODEL Q2N2955 PNP(IS=4.66E-12 BF=360 VAF=100 IKF=0.25 ISE=3.339E-11
+ BR=2 ISC=5E-9 RB=3 IRB=0.001 RBM=0.4 RC=0.04 CJE=5.802E-10 VJE=1.2
+ MJE=0.45 TF=8E-8 XTF=1 ITF=3 PTF=120 CJC=2.121E-10 MJC=0.4 TR=2.55E-6
+ XTB=1 )
Last edited:
I bought Intusoft's SPICE in the early 1980s, it was not cheap (~$500) and they brag about
their models. I have the files somewhere around here but I'd have to look:
SPICE Model Libraries
their models. I have the files somewhere around here but I'd have to look:
SPICE Model Libraries
Sounds intriguing.
Where can we see info about that stuff?
But this is not a single setup, so it requires a bunch of separate things to put it together. Likely build some of it, like a psu or whatever...
Runs on what platform?
I've been poking around their site, but with so much there, without knowing what exactly you were referring to, it's a little fuzzy.
I assume you were referring to those usb scopes, and the only software that I see associated with those things is what they call waveforms. So I guess it must be it.
But then there is the issue of platform, and as usual, I see windoze, and even linux, but no mac. Drats! always the same discrimination going on.
So I can't even look into it. I'm out.