Power Amp Stability Analysis in Older Amps

Jump to Latest
#21
I wrote an article on my website exploring why earlier amplifiers had compensation problems here

solid state amplifiers and feedback: a short history

it seems although we now take a lot of things for granted wrt amplifier stability and compensation, and especially after Self and Cordell, designers back in the 160s to early 1990s had to go through a learning curve.
 
  • Like
Reactions: anatech, Bruce Gillingham, Hipocrates and 1 other person
#23
The PL700 was an SOA nightmare from day 1. Until the 2SD555 came out there wasn’t even a viable output transistor available. It was the only Jap device that would do 200V rails. Delco PL909 was a POS, but Fairchild’s was close to a 2N5631 and only suitable for the 400. The D424 was very viable in the 400 but wouldn’t run at all in the 700. Ditto with selected 2N3773’s or 2N6259’s. There simply wasnt a single diffused transistor that would run with 200 volt rails. 160 yes, but 200 no - just a bridge too far. Then the D555 became unobtainium but Motorola had the 15024 by then. 4 ohms is marginal even with 21194’s but keep it reasonably cool and they will run. The 400 can be run HARD with 15024’s, 21194’s or D424’s (if you happened to have some).

But fans were always a good idea. I had a DJ “frenemy” back in the day that was too cool to use a fan on his Phase Linear, with results about like you would expect. Mine however suffered a far worse death when rivals disconnected the neutral at the panel of the venue. Remember about 2SD424’s not being able to run on 200 V supplies?
 
#25
It is exactly what it needs, but there is no simple way to do that. The QC triple is completely unstable on switched rails. I fully explored this back in like 2006 and gave up. I rebuilt the design, which originally had the PL output stage exactly as a fully complementary and it literally fired up and worked perfectly the first time. I still have the amp.

The other problem is the power supply. Unless you want to physically rewind it’s original transformer to get the necessary voltage taps you will never get a look-alike solution. No problem if you just intend on “building something”, but making a PL look original won’t happen.

A bucket list project that’s in the queue (and PCB set in design) is a close but not exact PL700-2 clone using 3 step class H and +/-115 volt rails. Same kind of heat sinking and chassis form factor, but internal transformers. Full comp of course to make it stable, and my usual modifications to the Series 2 circuit for stability. Meters are a similar functioning circuit I developed around LM3914’s. Complete with that pac-man clip indicator. The similarly “redesigned” BGW750 is further along, but both are back a ways in list of priority. Both would be “easy” to redesign around modern flatpacks - I already have a a PCB for a 2x360w/8R class H thats one single board and fits in 2U and runs off a pair of 500VA Anteks. I have parts to build 5 of them when my new shop can come on line. Will outperform PL700 in every way, but nowhere near as much fun.
 
#28
PB2 said:
Added Phase Linear 700 to the first post. Font size on my old LTSpice files seem to have
randomly shrunk when run under the new version. Anyone know of a simple was to fix
this without editing every one manually?
Click to expand...
I ran into this same problem when I upgraded to XVII.

From https://groups.io/g/LTspice/topic/font_size/73108438

I found and opened the .lib, .sym files and edited res.asy, npn.asy etc. They were .625 and I changed them to 1.5. Now it works!

Also this:

Years ago, the designer of LTspice added a new feature, the adjustable font size.
Unfortunately he did a bad choice for older schematics - he set the default for old schematics to the minimum size.

In the mean time, the designer implemented two options which allows to solve this problem.

Copy the icon of "LTspice XVII" on your desktop. Then right-mouse-clcik on it, choose properties and add the two options -FixUpSchematicFonts -FixUpSymbolFonts.

"C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe" -FixUpSchematicFonts -FixUpSymbolFonts

Start LTspiceXVII from this new icon.
Then open/drag the symbol files and schematic files with/to this LTspiceXVII. When you open the .asc-file, may be not all items are repaired (size 1.5) - then scroll the mouse-wheel once.
Finally save the symbol and schematic files.

And this:

Open LTspice from the command line with the -FixUpSchematicFonts switch. Any schematic witha dodgy font size you open will get fixed upon saving.

Easiest way to do this is:
  1. Make a desktop short-cut from C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe onto your Desktop
  2. Edit the short-cut: right-click -> Properties -> Target="C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe" -FixUpSchematicFonts ->OK
  3. Double-click on your new short-cut.
You can do the same if you have symbols with the same font problem, in which case step 2 should be:

Right-click -> Properties -> Target="C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe" -FixUpSymbolFonts ->OK
__________________________

I hope that helps.
 
  • Like
Reactions: bikeron and PB2
#29
Thank you, thank you, thank you, driving me nuts!

This works, the icon launches the program with this command, add the - options as shown:

"C:\Program Files\LTC\LTspiceXVII\XVIIx64.exe" -FixUpSchematicFonts -FixUpSymbolFonts

Open old .asc files by dropping them onto this icon, you might have to save it for the fix,
then reopen.
 
Last edited:
  • Like
Reactions: rongon
#31
kgrlee said:
😳 Are these mods & your 'full comp' detailed anywhere? These be da Holy Grail for me in da previous Millenium.
Click to expand...
The full comp mod was used on later production units when the MJ15025 became available. It just replaces that negative half with an EF3, and adjust the bias range to suit. The BIG advantage to doing this is the increased local stability. The amp is stable at ANY quiescent bias which the heat sinking can handle. No more nasty crossover distortion or jigging with it to rid of pesky local oscillation - all you have to do is put 3mA per output transistor. It’s technically underbiased but light years ahead of original settings. One can run at higher bias and not oscillate, but the heat sinking won’t support tons of bias. Once able to run with a bit of bias in the output stage, you can then return the bootstrap cap for the VAS to the output instead of the driver’s emitter. If you’ve ever had a leaky bootstrap cap you’ll know why this is a good idea.

Other updates: Add 100uf of bypass capacitor on the rails of the driver board. Good idea regardless. Solves many phantom stability problems. Reduce the lead compensation cap (from VAS output to op amp summing node) to 27 pF. 56 pF has been recommended by others. Original value is wrong. Add local feedback to the emitter of the MPSA93 that sits between the op amp and 40412. Replace the 1.5k resistor with 150 ohms, and add 10k 1 watt from the output to the junction of the resistor and MPSA93’s emitter. Add reverse bias protection on the MPSA93 (just a 1N4148).
 
#33
The QC side liked to break into local oscillations at bias currents that you would normally want to run the amp at. That was “solved” by running fully class B with no bias (vbe = 0.35V). It was highly dependent on which output/drivers were used. I could get a little bias in 2SD424’s on a 400. Not with anything else. And nothing on a 700 - even MJ15024’s would produce non-destructive snivets. PL 909’s, however, could explode. Use PNPs in the bottom, and the output stage is trouble free. Of course there were other stability problems too - especially in the S2 where he tried to get away with using many of the same compensation values as the original…. with far more loop gain and an extra pole.

I could NEVER stabilize the triple QC against the fast rising edges of class H rail switching. I tried to build an “all 2N3773“ class H on +/-93V. They were the modern Motorola version, just before the ON switch. I tried absolutely everything over about a years tinkering. Finally just said fire truck it and jumpered the board to take a batch of MJ15016. Just turned it on with the dim bulb and I had sound. Ran flawlessly.

Down on lower voltage, I couldn’t make the QC triple oscillate. I built basically a “Phase Linear 60” with +/-28V rails and 3 pair of 2N3055’s, with TIP drivers. Biased optimally (26 mV across emitter resistors). It actually sounded damn good and drove 2 ohm loads easily. Stable as all hell even driving a capacitor directly.
 
#36
No, when I started doing full comp mods the B600 was long gone, and MJ15024/5 were used exclusively.

The “KM441” was essentially a full comp PL400S2 with some optimization and a lower +/-60V supply. Same physical size trafo (identical lam stack). It used D424/B554’s and those uber-fast 80 MHz Toshiba flatpacks (C2565/A1095) for drivers. It had all the other stability mods plus a current source loaded VAS. I didn’t own a 700 at the time, and I had two KM441’s and ran them bridged to run gigs with. I was having a little BOC concert in my dorm room at university with them on Halloween, and managed fry a pair of 18” Pyle drivers with them. We eventually did figure out what the smell was.

The only older model I have is a 2N3773. And I doubt I could find it right away either. The machine I had spice on has been decommissioned, and all my media is still all packed away.
 
Last edited:
  • Like
Reactions: PB2
#37
I saw a few Phase Linears in a high end audio emporium in the late 1970s along with the joystick preamp. Huge beasts that sounded impressive but given everything I know about them now, and the little bit about amp design, I’d say they were a bridge too far and with many fundamental issues. Luckily nowadays we have LTspice and similar that make compensation more straightforward and fantastic, cheap output devices. But, there is also a lot of good documentation and ‘working knowledge’ on this stuff thanks to Self, Cordell et al that was not around back then. Most designs as a result were done on a wing and a prayer.
 
#39
QC was still used for high voltages until the mJ15025 came out - that was like 1985. NPN was released by itself a few years prior. Jap manufacturers were there first - but Carver didnt start using B600’s - probably because they were expensive as hell. Never saw a 400 manufactured with D424/B554 (or even all D424 either) but it would have been eminently practical as they were cheaper than Motorola, and the PNP was maybe $1 more instead of double. One could have also voltage-selected MJ15004’s (or 003’s) for use in a PL400 also, but never heard of it being done. Fairchild FPL909 was essentially that, because it was Fairchild’s 2N5631 which WAS better than Motorola (Royally pissed them off too I’m sure, as that was the heyday of the rivalry). Moto 2N5631’s would survive in a DC300A, but not a PL400. Later vintage restorers and other fanatics didn’t play games with voltage ratings, they just used appropriate parts. Yeah, there were cheapskates fixing them with selected 2N3773’s as I did before I discovered MCM and could get D424’s, but modern restorers wouldn’t do that. By the time these amps became a cult icon, one could get MJ15025’s any day of the week - even without at tax ID. AND they were the best parts for the 700 to boot.

I think PL boiled down to “how cheap CAN we make a 700W amp?“ And its baby brothers. And it wasn’t expected to handle idiotic things like running more than one pair of speakers in parallel or taking them on the road. But people (with little money, such as myself at the time) did.
 
Last edited:
#40
I don't follow digital power amp or power supply design, but I would think that a PWM power tracking
regulator that keeps the voltage cross the output transistors say 20-30V above the output signal
would be easy to design. Yes, it would have to track say to 50 KHz but if a digital amp can be built
why not? Take the +/- 100V supplies in the PL700 and lower them with a PWM regulator.
 
You must log in or register to reply here.
Top Bottom