Old Dynair line amps: ideas for improvement

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

NewYorkDave

Well-known member
Joined
Jun 4, 2004
Messages
4,378
Location
New York (Hudson Valley)
There was a thread here on Prodigy about these amps, including some great analysis by PRR, but the thread seems to have disappeared. Bummer.

Schematic

I rescued several of these from the scrap heap. I was going to take the transformers and junk the rest, but I'm kind of reluctant because they're nicely-made amps (and, for what it's worth, apparently rare).

Looking at the circuit (which I traced myself, no schematic was available) I see a couple of areas for improvement. First is that terrible input arrangement, with the bias current for the transistors being drawn entirely through the wipers of the level pot. (As you would expect, this caused a lot of noisy if not downright jarring intermittents when the amps were still in service). The step-down ratio of the input transformer seems a bit excessive except for the fact that the pot value is so low--and I might get rid of that pot, anyway. Beefing up R11 (higher power rating) and C1, C2, C3 (higher voltage) and increasing B+ by a few volts is an obvious thing to try.

Bypassing R11 might be worthwhile; the net signal voltage across is is zero in theory but that's only if the output pair is perfectly balanced (never). My experiments with push-pull outputs in tube line amps showed me that whether the cathode resistor was bypassed or not could have a measurable effect on the distortion content--with more higher-order distortion products with cathodes unbypassed, at least in the circuit I was working on.

It does seem odd that the DC operating points are arranged so that R11 has to be so big; half the supply voltage is dropped across it. But as I recall PRR pointing out, this does make the output section hard to blow-up.
 
Well, I took a break from work this afternoon, grabbed one of the amplifiers and opened it up on the bench. I uprated the lytics and R11, powered it off 15V and checked to make sure everything was OK. Then I slowly ramped up the B+ to 24V.

An odd thing happened... when the B+ rose to a certain point, the output transistors started going toward cutoff. Huh? Quickly, I realized what was going on: the base bias for Q1 and Q2 off the resistive voltage divider R1 and R2 was riding up with the B+ and biasing Q1/Q2 into heavier conduction, in turn causing a greater voltage drop across R7/R8 (dropping more sharply than the B+ was rising). This, of course, caused the collector current of Q3/Q4 to drop due to the reduction in their base bias. So I reworked the base bias for Q1/Q2 and everything fell in line.

With a 24V supply, the amp can swing +27dBM into 600 before the onset of clipping. That's all well and good, but I don't think a simple increase of B+ is the way to go with this circuit. A lot of power is wasted in R11; Q3/Q4 are drawing almost 120mA (!) and 1.4 watts is wasted just heating up that resistor. Q3/Q4 get very hot, too. (Ya figure about 12V dropped across each at 60mA collector current, that's 3/4 watt right there, too much for the 2N697).

I think a redesign might be a better approach. The first thing that comes to mind is to increase the value of R7/R8 (reduce base voltage and collector current of Q3/Q4 and also increase gain of Q1/Q2) and reduce R11 accordingly.
 
It looks like an old tube circuit converted to work with that new fangled solid state stuff..

The two power transistors are biased slightly on kind of like Class AB. The positive half waveform is provided by one driver the negative half the other... Apparently looking for 15v rail so when you powered it up higher the too hot bias was like the amp trying to swing pos and neg at the same time.

I see little reason to pursue this old topology other than academic interest, or an interesting distortion source.

JR.
 
little reason to pursue this old topology other than...

...the fact that I have about 15 or 20 of these amplifiers :wink:

It looks like an old tube circuit converted to work with that new fangled solid state stuff..

That jibes with the amplifier's time frame. Three years ago, I emailed Dynair's founder asking if he had any info on these amplifiers... Lazy-assed me, just trying to save myself the trouble of tracing out the cicuit! :wink:
He replied as follows:

The AD-3081A and the PS-3015A were developed in 1963 at the same time as the 5100 switching equipment was designed.

When I sold the Company in 1995 I did not bring any instruction manuals or schematics with me.

Since the products were, I believe, designed entirely with discrete components and single-sided circuit boards, the circuits should be traceable without too much difficulty.

I'm sorry that I can't give you any real information.

You made my day with the news that your System 21 is still in service after 22 years!

Garry Gramman

Only three days later, he passed away at the age of 82 while fixing a noisy fan in his attack--a true broadcast engineer to the end.
 
Only three days later, he passed away at the age of 82 while fixing a noisy fan

:shock:

Times like this remind me that I would rather go out when doing something for myself rather than stuck in a hospital bed with someone having to care for me.

Sorry to go OT, back to the circuit.
 
By the way, the System 21 routing switcher mentioned in the email is still plugging away--they just don't make 'em like that anymore! Let's just say our infrastructure upgrades are running a couple of years behind schedule.

As for the line amps, I'd be happy if I could just tweak them to get a clean +24dBM out. Sure, they're usable as-is, but whenever possible, I like to maintain 20dB of headroom above standard operating level in my systems.

The 5K symmetry trim could be relabeled as a distortion control...

You could certainly make it a user-accesible control if desired, 'cause it doesn't have nearly enough range to do damage to the circuit.
 
Finally! Another symetrical SS amp superficially resembling the Altec 9475A and Langevin AM-16. I knew there had to be another version or two out there. Can we see a pic of a module?
 
Well, it resembles them insofar as it's push-pull, xfmr-coupled in and out, but the Altec and Langevin are a bit more complex (and use special xfmrs, IIRC, at least in the case of the Langevn). I don't have those schematics at hand at the moment

I think I took a couple of pictures a while back. I'll look for them.
 
By that I mean there's nothing else out there (in US products) that's PP beginning to end, period, in SS low level amps. There's a SS RCA talkback amp which is PP and has interstage transformers between every stage. That's the next closest thing I've come across.
 
Hey, whatta ya know... I remembered something I'd seen in another Dynair manual (since thrown away by an overzealous co-worker :mad:) for video amplifiers that plugged into what appeared to be the same frame. The supply voltage on those was +18V. I measured +15V on the supplies I have here, but it's possible that they may have drifted low after 30+ years. The PSU circuit is old school, all discrete, with trimmers and whatnot. Anyway, I decided to try +18V.

Everything fell right into place with no mods to the circuit. Max output (just at the onset of clipping) rises to +25dBM while at the same time, Q3/Q4's collector current stays about the same, therefore they dissipate about 330mW each, and R11 dissipates no more power than it did before. Output at +24dBM looks clean--haven't done a THD measurement yet. A spot-check of frequency response gives a so-so result, -1dB at 20Hz and 10kHz, -3dB at 15Hz and 17kHz, a whopping -4dB at 20kHz, all from a generator source impedance of 50 ohms. (I'm gonna double-check my test setup when I have time. The plummeting response above 10kHz doesn't seem right).

So I'm strongly suspecting that these things may have been intended to run off 18V, although that doesn't explain the 20V rating on C1 and C2 :?

+18V supply:
Q1/Q2 Emitter: 0.23V, Base: 0.89V, Collector: 7.7V
Q3/Q4 Emitter: 7V, Base: 7.7V, Collector: 17.4V

EDIT: Corrected typo.
 
Hey Dave,
These amps look awfully similar to my Flickinger amps.
I am running them at +-18V, and a minor change in the voltage induces buzz. I actually trimmed the supply voltage to hae the cleanest audio without input.
PRR gave me some cool info on how to get them up and running to the mic amp level, but I can't find the original thread. I will have to find my printouts and scan them if you'd like.
 
Found the datasheet for the 2N697. It certainly wouldn't be my choice for Q1/Q2, although it's not bad for Q3/Q4. I guess there weren't that many choices 44 years ago, or the designer wanted to use the same type throughout for economy of scale.

I think using a proper small signal transistor for Q1/Q2 could be a worthwhile mod.
 
[quote author="NewYorkDave"] 2N697. It certainly wouldn't be my choice for Q1/Q2, although it's not bad for Q3/Q4. I guess there weren't that many choices 44 years ago, or the designer wanted to use the same type throughout for economy of scale.

I think using a proper small signal transistor for Q1/Q2 could be a worthwhile mod.[/quote]


Sounds like a good idea. They probably were thinking economy of scale, along with 'it's only 10 db, doesn't matter'. They did have quite a few choices by that point. I'd guess it was all so 'newer/better' that the 'latest' might easily always have been the choice.
 
I would use output transformers only.

Or probably input transformers too... Are they for 600 Ohm input?
 
> The two power transistors are biased slightly on kind of like Class AB.

Can't (shouldn't) run AB. The resistance in the emitter is 12R for class A, 112R for class B, big shift of gain, massive distortion.

Let's see: 68mA total bias current, thrown to one side of a 500CT winding, is 125*0.068= 8.5V peak in Class A. That's roughly all the voltage we have. It can run class A. Max undistorted sine power is around 0.289 Watts or +24.6dBm. Not shabby.

We really need 16V supply to get 8.5V peak swing.

ASSuming 15V supply: Transistor dissipation is 0.212 Watts each, which is plenty for older packages even with fins. (And we are down to +21.7dBm, oh well.)

If you think the transistors will stand 0.37 watts dissipation, apply 19V, get 0.057A peak current, 12.8V peak voltage. This will drive a 900 ohm load to 25.6dBm, only 23.2dBm in 600 ohms.

Ah: voltage rises, current drops, the ratio V/I soars. The ideal voltage for 500CT is 16.66V. At 15V you "should" load in 306 ohms, at 18V it likes 708 ohms, at 20V you should load with 1.2K for best power output. The max power output does not change a lot with supply voltage; the optimum load changes very fast (300@15V, 1200@20V).

They were struggling with new technology. (And possibly hampered by patents to avoid.)

The low input impedance would be typical of low-noise Germanium design. Input leakahe current was horrid. But this is Silicon; they could have aimed higher. Especially with a gain of just 40. Suspect that T1 was the only super-hi-Z winding they could source cheaply, and they just accepted the 500 ohms.

It NEEDS a regulated 16V-17V supply, or a re-think. You can't complain about a "flawed" design which served well for 40 years and half a million hours of program, but there are other ways to skin this cat.

Drop supply to 12V. Pick R11 to drop 0.76V at reasonable current. Lose R1 R2 and connect R11 there (or to new C-R base networks if you hate DC in the pot). Say you swing 10V in 125 ohms, 80mA peak, 40mA per side, 80mA in R11, 0.76V/0.080A= 10 ohms neer-enuff. Each transistor dissipates 0.44 Watts, which may be just-OK for the old TO-5 with fins.

**** oh, drat. That gets the total current under control, but the differential current is hyper-sensitive to Q1 Q2 balance, which may be bad. But no worse than before? So VR2 may pull it in. How bad did the balance drift over 40 years?

I'd keep the antique transistors. You can get better specs today, but this is a simple chore. Also, J. Lindsey Hood used them.

> I see little reason to pursue this old topology other than academic interest

No need to get rude. Anyhow, academics never go near stuff like this, and wouldn't understand if they did.
 
Thank you PRR. I see several things that were obscure before, and more with some re-reads.
 
Ran the amp (still with +18V supply) on our recently-acquired Audio Precision.

All measurements were done with 620-ohm load. EDIT: Interesting thought: when you factor output transformer winding resistances into the mix, the load seen by the output devices turns out to be right around what PRR calculated as optimal for a 18V supply.

+24BM output:
THD+N (10Hz-300kHz, unweighted): 0.75% at 1kHz (rises to 1% at 40Hz)
IMD: 2%
Noise: 145uV

+14dBM output:
THD+N (as above): .03% at 1kHz (rises to .06% at 40Hz)
IMD: .12%

Not bad.

Frequency response flattens out at the lower output level... Ref. my earlier observation about the falling HF response at +24bBM output.
 
So, with an eye toward reworking the input network, I wanted to calculate the impedance looking into the base of Q1 and Q2. I'll run through it here; please let me know if you spot any errors.

Here are those voltages again:
+18V supply:
Q1/Q2 Emitter: 0.23V, Base: 0.89V, Collector: 7.7V
Q3/Q4 Emitter: 7V, Base: 7.7V, Collector: 17.4V

I figured total Q1+Q2 base current to be 0.135mA, which is 0.0675mA per transistor--assuming it divides equally, which is a big assumption! :wink: But the net result is the same since we're concerned with differential input impedance and could treat it like "one big transistor" for the purpose of these calculations. But I'll do it per-transistor:

HFE = Ic/Ib = 3.12mA/0.0675mA = 46.22

RIE = 25/3.12mA = 8 ohms
RE = 81 ohms

ZIN = (81+8)*46.22 = 4.1K, therefore Diff. ZIN = 8.2K

Kinda low, but if you dispense with the pot and use a 10K:10K input transformer, it's reasonable. Problem with getting rid of the pot is that you don't always need ~30dB of gain. There's also the option of using a stepdown input xfmr, just with a lower turns ratio than stock.
 
Back
Top