High Power Low Distortion Amplifier with TPC

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I came across this schematic (2 of them, actually) along with measurement charts from the product manuals in my recent search for a new (to me) amplifier. Detailed measurements for amplifiers with schematics actually available are far more difficult to find than they ought to be. Proper service manuals or even schematics seem to have fallen out of favor, and measurements have never been in favor. Anyhow, for a commercial product, I think the performance of this amp is pretty amazing, and felt it worth sharing. It’s always nice to see a few advanced ideas actually at work in a commercial product, and to see how it performs.

Measured THD+N of this amp is under .008% at 20kHz into 8 ohms at 425W. Back off the power by half, and THD20 drops to around .0045%. For a one-off DIY amp, this might not be that remarkable. But for a commercial product? It’s unusual. If you dig into various measurement archives such as those at Stereophile or Soundstage, you quickly realize that for all that is known about building amps with minimal distortion, little is ever done in the commercial market. If amplifiers were reviewed like cars, the reviews would say that most of them were broken and needed to go back to the drawing board.

What makes this really cool is that this is professional amp that can be had for a few hundred bucks used. And just look at the schematic. It’s full of stuff out of the diyaudio textbook. Boosted rails? Check. Beta enhanced VAS? Check. Cascoded differential inputs? Check. Baker clamps and a soft clip circuit? Sure. “Advanced” Compensation? Yup. As far as I can tell, the amp uses a two pole scheme. Although the time he wrote it is not clear, Doug Self said that TPC was “rarely applied to power amplifiers.” This was done nearly twenty years ago across tens of thousands of units.

FWIW, the amp for which the performance chart is posted is a Mackie 2600. Schematics are posted for the 1400 and 2600, both from around 1997/8. The specification on the 1400 was even better than the 2600, so I would not be surprised if the measured performance was even better. The 1400 is the one with the differential VAS. It’s also fair to note that these measurements presumably include the “preamp” section which uses a pretty big handful of opamp buffers for a volume pot, a subwoofer crossover, and a high pass filter. Most of this crap is in the signal path at all times, which makes the measurements all the more remarkable (and probably helps explain the noise...). I’m fairly sure you could just wire up a set of RCA jacks to bypass all of that crap. The amp section seems like a complete package.

Anyhow, hopefully this is useful to someone. It seems like there’s a lot of good stuff here. Any thoughts about these designs? Areas for improvement?
The circuit descriptions in the service manual go on for pages and pages, and I could post chunks if anyone is interested.
 

Attachments

  • 2600FE.jpg
    2600FE.jpg
    217.2 KB · Views: 902
  • 2600THDvFreq.jpg
    2600THDvFreq.jpg
    73.4 KB · Views: 883
  • 1400Sch.jpg
    1400Sch.jpg
    201 KB · Views: 891
  • 2600THDvOutput.jpg
    2600THDvOutput.jpg
    91.5 KB · Views: 868
I was playing around with this amp today and it reminded me of this thread. Didn't get a lot of attention the first time around, so I thought I'd bump this up a year later to see if it generates any interest since I think this is still interesting. Possibly a good platform to do some modifications and have a really solid amp. I've not seen a review of anything that comes close to performing like this for anywhere near the price.

So... Are these really high powered two-pole compensated amps with miniscule (for commercial products) amounts of distortion? Anyone care to dissect what's going on with this and how they managed to squeeze (for a commercial product) vanishingly low THD20 out of a PA amp?

You can dig through hundreds of published reviews and measurements, and you won't find nearly anything this well behaves this well on the high end for anything close to this price. As far as I can tell, this design was completely in-house by Mackie and "original" (so far as anything in audio is)...
 
It’s also fair to note that these measurements presumably include the “preamp” section which uses a pretty big handful of opamp buffers for a volume pot, a subwoofer crossover, and a high pass filter. Most of this crap is in the signal path at all times, which makes the measurements all the more remarkable (and probably helps explain the noise...).
Actually it looks like almost all of this can be bypassed:
Limiter / Sub EQ: Off
Amp Mode: Stereo
Jumper J8 (Mode): LAB-REF

This leaves little more than the balanced input stage and the output level circuit.

Hmm, I think I just found a problem. The output level circuit must have been designed by the intern or something. A 5k pot (when the input stage is sending up to +22 dBu into there), a 2k2 input resistor... and an NJM4560 (~10 nV/sqrt(Hz)) used as a buffer. Which I guess isn't too fussy about local bypassing, which consists of skimpy 10n caps. Still, it seems a bit silly to be loading the input stage that much (~3k3, about the minimum of what I'd subject an NJM2068 to at these levels), only to follow it up with a relatively noisy part. Bit of a waste. Not a super big deal at 32.7 dB of gain, mind you, I'd guesstimate maybe 50-60 µV of output noise or something. Still, it could be better.

This amp normally should not be noisy at all... assuming you are running pro level into it, that is, which nominally is 12 dB higher than consumer. If you need to keep input level fully cranked up, expect almost 5 µV coming in from the balanced input stage (which is running at unity gain), so that's somewhere around 200 µV there. Still not super-duper noisy though, more like run of the mill 85 Wpc @ 100 dB SNR integrated territory.

Oh, and add whatever your preamp puts out to those 5 µV. Plus some more noise from the highpass filter if left in.

If you are getting ground loop issues (this is an IEC Class I device after all, as pro gear tends to be), then all bets are off, of course. Off-the-shelf RCA to XLR cabling tends to be 100% unbalanced, completely wasting any and all advantages of the balanced input stage. Either roll your own or use a line isolator, or even consider modding your preamp with an impedance-balanced output (which takes just a handful of passives to replicate output impedance between cold and ground).

If there is audible noise with the input level turned down and the speakers don't happen to be horns, well, yes, that would be a problem.

As for power amp modding ideas, I see that feedback is returned to input ground, a common but arguably debatable choice. The "A" ground in the board may not be suitable, one might have to run a dedicated connection to the output terminals (but watch out for stray inductance and coupling).
It also looks like the cascodes are referenced to +/-15 V (with some minimal LC filtering), which are generated by a pair of 7815/7915 regs (not even 78M series parts which tend to be lower in noise... I can't imagine they're seriously using anywhere close to 500 mA, let alone 1.5 A). Granted, the cascode shouldn't care that much, but attenuation is definitely going to be finite, and standard issue 3-terminal regs can be quite noisy.

The distortion figures aren't bad and probably conservative, but something with 8 pairs of output transistors should be able to do better. Very good amps easily add another zero these days.
 
Actually it looks like almost all of this can be bypassed:
This amp normally should not be noisy at all...

Thanks for the input! Same conclusions I came to regarding the stuff on the inputs: Most of the preamp circuitry (including the input buffer, crossover, input attenuator, etc.) should be easy to bypass. My recollection from when I looked over it is that you should be able to wire a set of RCA jacks right to the AMP-IN pin and be okay. You must have pulled out the big schematics... I don't think the ones I posted showed that test pin jumper. ;)

The distortion figures aren't bad and probably conservative, but something with 8 pairs of output transistors should be able to do better. Very good amps easily add another zero these days.

Ignore the "brochure" numbers. Look at the measurements I posted. That's where the magic is. Another zero isn't added so easily. The THD20 for the Mackie is .0045% at 1/3 power, measured at the time on an old AP System 1. .00045% THD20 would be better than a Benchmark AHB2, which is about as good as it has gotten for a commercial product to date. Stereophile still has about the only decent amplifier measurement database out there that tells you much that is useful. There are precious few amplifiers this well behaved in their database. Take the Classe CT-M600. At 50W into 8R, the THD20 is at .007%. With the Mackie the amp is at the same distortion at full power. At 25W a Bryston 7BSST2 is at .01%. At 1/3 power the Mackie is half the THD of the Bryston. A Rotel RB-1090, which is still $1000 used, is up to .025% by 100W at 1kHz. The Mackie is nearly a full order of magnitude lower. Boulder, Benchmark, Bryston, and a handful of $50,000 amplifiers are the only products I've seen measured that approach this level of THD20.

It was those plots in the back of the manual that made me take another look at this thing. 20+ years ago Mackie was turning out amplifiers which had incredible linearity. And I'm fairly certainly they did it using a compensation scheme no one else had previously used in a commercial product, and which precious few have used since. I know Hypex amps are all the rage, but why not grab an old Mackie for $150 and play with that? The basic circuit is possibly (in terms of linearity for a high power AB amplifier) among the best anyone has ever soldered onto a PCB, and sold by the truckload. Fair?
 
Last edited:
My recollection from when I looked over it is that you should be able to wire a set of RCA jacks right to the AMP-IN pin and be okay.
As far as unbalanced inputs on an IEC Class I device go, anyway... a combination that I'm not particularly fond of. You'll get ground loop problems faster than you can say "ground loop". Be prepared to be jumping through some hoops when trying to connect a PC.

No, this guy should get the balanced (semi-)pro audio interface or mixer output it deserves. Some juggling of output level vs. input gain will be required to obtain optimum SNR. I'd try aiming for +10..+16 dBu peak as a reasonable compromise, requiring an input gain setting of 27 to 21 dB for full output power (and correspondingly less for lesser needs).

I might also consider using something lower noise for the output level buffer opamp. I imagine NE5532 or LM833 (the ex-NatSemi part with the complementary output stage, not the TI one) might do, maybe with some more bypass capacitance. Unfortunately the amp is pretty heavy on surface mount, but that's not quite as daunting to the hobbyist as it used to be.
You must have pulled out the big schematics... I don't think the ones I posted showed that test pin jumper. ;)
The service manual I found at first didn't contain any schematics, so I had to do some digging.
Ignore the "brochure" numbers. Look at the measurements I posted. That's where the magic is. Another zero isn't added so easily. The THD20 for the Mackie is .0045% at 1/3 power, measured at the time on an old AP System 1. .00045% THD20 would be better than a Benchmark AHB2, which is about as good as it has gotten for a commercial product to date.
It's definitely good enough to be considered "blameless" at this point, and certainly excellent by the standards of classic PA amps. That said, commercial amps have been able to do at least as well since the mid-'80s. For example, a Yamaha M-85, their TOTL model back then:
Minimum RMS power per channel, 20 Hz - 20 kHz:
0.003% THD 8 ohms: 260 W
0.007% THD 8 ohms: 295 W
That's for the US/Canada model, 220-240V regions apparently got a beefier transformer and were spec'd at 340 W each.

Even the M-2 from the late '70s was spec'd at 240 + 240 W, 8 ohms, 0.005% THD, 20 Hz - 30 (!) kHz. Unsurprisingly, Yamaha's PA amps from that time also enjoy a reputation of being nearly bulletproof, high quality amps well worth restoring.

Another Japanese amp from the early '80s, the Kenwood Basic M2:
THD, 20 Hz - 20 kHz
0.004% at rated power and 1/2 rated power, 8 ohms
Rated power is 220 W @ 8 ohms.
Same specs for the M2A.
Even the lesser M1A wasn't bad: 110 Wpc, 8 ohms, 20 Hz - 20 kHz, THD 0.004%.
THD at 1/2 rated power: 0.003%.

So performance figures in this range, while good, are not at all outlandish. This was SOTA territory almost 40 years ago. Distortion approaching ppm level (0.000x%) is doable these days but few people actually care, which is why you rarely see it in commercial designs.

It was those plots in the back of the manual that made me take another look at this thing. 20+ years ago Mackie was turning out amplifiers which had incredible linearity. And I'm fairly certainly they did it using a compensation scheme no one else had previously used in a commercial product, and which precious few have used since. I know Hypex amps are all the rage, but why not grab an old Mackie for $150 and play with that? The basic circuit is possibly (in terms of linearity for a high power AB amplifier) among the best anyone has ever soldered onto a PCB, and sold by the truckload. Fair?
Fair enough. In DIY, that sort of money barely gets you a decent case - as so often, prebuilt is very hard to beat on price (though at 55 lbs / 25 kg, shipping is likely to cost an arm and a leg). The PA guys are probably happy to get rid of their heavy AB dinosaurs at this point, and if you have any need for an amplifier in this class and operate it infrequently enough for it not to make much of a dent in your power bill (the output stage needs a nominal 91 W secondary-side already, so I imagine the whole thing will be drawing something close to 150 W from the mains in idle), you could probably do an awful lot worse. 2x 425 W @ 8 ohms or even 700 W @ 4 ohms still is a ton of power by most people's standards.

How's the fan noise? I imagine a well-used unit would appreciate at least a good cleaning at this point; the odd fan bearing might need some attention as well.
 
Last edited:
Anyhow, hopefully this is useful to someone. It seems like there’s a lot of good stuff here. Any thoughts about these designs? Areas for improvement? The circuit descriptions in the service manual go on for pages and pages, and I could post chunks if anyone is interested.

2 stage differential (symmetric version of the old sansui topology). The VAS itself is the 2'nd differential . LF loop gain must be 135db+ .

Would be an easy "slewmaster" input stage. As far as simulation , I would just have to replace the hawksford VAS on my leach type input stage with this one.

No wonder it does 1/2ppm , gain margin @ 20k must be way above average for a VFA.

EDIT - Yikes , they are just driving an (ancient high Cob) EF2 , my triple OPS would at least halve the THD.. Yup , I would tie all the boosted beta VAS collectors to the references of a added cascode. Then I could use BC550/560 and 2sa1381/sc3503 for the cascode. .

OS
 
Last edited:
you got me going , I thought of it while designing to 20+ slewmasters , but a cascoded hawksford (super pair) .. i never got around to it.

Slightly less gain than the beta enhanced differential on this amp. But (no baker clamps) ... superpair is self clamping. I increased the gain of the first stage. Boom 140db OLG LF + rounded "tubelike" clipping.

I simmed both this amp and the superpair , real close with 1ppm 20k....

Super pairs pole is very high , some of my fastest CFA's would be slighty unstable with it.

Actually had an oscillator with a sansui 2 stage differential driving a cascoded superpair. 2 stage differentials are notorious as being difficult to compensate. lead and lag , 2-3 FB paths or miller compensations. I like either a single shunt or a small miller (pair- comp). I will show this amp + the superpair. Both amps can match or exceed a doug self "blameless" with out TMC.

OS
 
you got me going , I thought of it while designing to 20+ slewmasters , but a cascoded hawksford (super pair) .. i never got around to it.

I simmed both this amp and the superpair , real close with 1ppm 20k....

I will show this amp + the superpair. Both amps can match or exceed a doug self "blameless" with out TMC.

OS

So something useful is starting to come out of this... :) (I'd forgotten about the old Japanese amps sgrossklass mentioned as particularly low distortion designs. Well, half forgot and half disregarded: The prices on the Yamahas in particular are through the roof for a pc4002 at USD $1000+, most are 15 years older than the Mackie, and they impossible to find.. even the pc2002 is $600+) The Mackie is a US-built PA amp generally written off as a piece of old junk that is below a Crown or a Crest. It's not. It is far more advanced than most US "high end" amplifiers. It should be the equal of a PC4002 (or better) for about $350, and a PC2002 for $150.

So far as the output stage, yes, it's just a standard EFT, but it does have boosted rails, at least, and will handle any load with ease. Simple, but far more fancy than you'll usually find in a "PA" amp. The front end is completely without compare in ANY other affordable high powered amp.

Anything you see in the schematics that looks particularly problematic, or that should be replaced as a simple improvement? You've got far, far more technical knowledge than I could hope to have. Obviously, bypassing all of the input/preamp garbage and running an RCA jack directly to the AMP IN pins is top of the list (balanced is nice, but let's be realistic--most preamps aren't). But beyond that?
 
... in D. Selfs 6th edition of "Audio Power Amplifier Design" the "Blameless" has now TMC included (D. Self calls it "combining two pole and output inclusive compensation"). Hard to beat!

Have fun, Toni

Easy to match without baker clamps and TMC. Easy to beat even with the simplicity of a CFA and a fancy VAS. (Or a OLG "monster" like this amp or the old SANSUI "el cheapo" 2 stage diff. design.

Old 71' design does 10ppm or better with 21st century components - original was .01 - .02 thc 20k.) Wolverine (brother to badger) , did a real world 5ppm @20k 250w with the ef3 in the slewmaster thread. Vzaichenco's Russian designs all average 2-10ppm with better even/ odd products , as well. BTW he went on to non-switching AB + EC OPS's ,I am still primitive with my Harmon Kandon (improved) EF3's.

Still , 1-10ppm at massive powers (without special comp.) Then a special Groner front end - simmed at sub ppm.

Many could have TMC applied , the russian has ppb front ends. Difference is , me and the Russian have native non-saturating ips's. Run the OPS 3-4v higher behind a cap multiplier - good clip to vacuum clip with even harmonics ... super-pair hawksford does this (hawksford 2 x red led's will blink for clip - neato.

DIYA's "badger" 20ish ppm non-tmc - maybe 5-10 with a lightly loaded EF2. Best documented topolgy in audioland - That is why I chose it.

I would like to see the physical layout of rmhuizenga's amp. All ppm amps only get down below 20ppm with meticulous ground returns and other layout "tricks".

OS
 
Last edited:
Different design goals now.

I would use use a variant of this for an IPS. But SMPS + SMPS modulated class G/AB OR ... SMPS + 2 X bridged 40 - 0-40V AB = the SOA 'breakpoint". Way more SOA , low Heat.

AB X2 = less heat , more efficient than 80-0-80 rail AB.

SMPS - modulated G = 95% PS + almost 80% OPS , approaches class D. Earth ain't getting no cooler. Got SS ab's with valve like clipping , just lacking the nostalgia of the warm glowin' sovtek valve and the big blocks of earth rapin' copper an iron.

OS V2
 
I would like to see the physical layout of rmhuizenga's amp. All ppm amps only get down below 20ppm with meticulous ground returns and other layout "tricks".
OS

The layout is a little complicated since they used double sided PCBs and lots of SMD. I snipped it and attached it. This is for the m2600. I seem to recall that the m1400 uses a single PCB for the entire amplifer from input to output--both channels.

Looks like most of these were probably built around 1999 to early 2000s, judging from the copyright dates. That makes them a lot newer than the old Japanese amps. As common as it is to see THD20 at .00X here on DIYA, it's easy to forget that it is very rare in commercial products. Most "high end" amplifiers are still using designs that would have been current in 1980.

Also, I forgot to mention that I did replace the fan with a Panaflow. Since it has a little "kicker" to give it some extra juice at startup, it works at very low speeds and with zero noise. Let it idle for days sitting by my desk after I forgot it was on with no problems. For home use, if it idles fine, it will probably be fine in operation. I haven't gotten around to bypassing the the "preamp" stuff, but I'm not sure it's really necessary.
 
I attached the layout but now it isn't showing up. Trying again... This is the input stage and not the output. If you want any of the other schematics or service descriptions let me know and I can post them. It's not too easy to find the full manual that has everything in it.
 

Attachments

  • m2600 layout.jpg
    m2600 layout.jpg
    370 KB · Views: 178
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.