|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
130.203.167.222
In Reply to: Now we are talking... posted by serus on September 29, 2004 at 19:30:31:
First things first, I measured only the front end stages, loaded lighter than they would have been had the power valves grids been involved. No loop nfb, and only the gain and P-S stage was being put through its paces. NFB cleans the whole amp up nicely and to the limits of what the OPT is capable of. So flippin' what...I found no measurable un-balance with CCS cathode load diff amps. i don't give them any NFB input and I only would publish test results which I had duplicated with good tubes, known good and NOS, and incidently some weak sounding used did still measure quite well. Good matching makes it very difficult to determine which is which at a first glance. Two identical traces...180-degrees OoP
Split load inverter is only going to have a maximum output of 25% of it's B+, and likely less when it is part of a direct coupled circuit. Between a fifth and sixth is allIcould tweak out of mine, and at published Dynaco OP points, it is more like a sixth. which for a UL rigged outut valve is just enough.
The NFB-less triode V-A and direct coupled split-load did not ever measure well, usually not even close to acceptable. More like POOR, sometimes *REALLY* poorly.
regards,
Douglas
!
Beware the man who behaves deceitfully towards others, for there is nothing else to save you from his deceiving you
Follow Ups:
> First things first, I measured only the front end stages...
> I found no measurable un-balance with CCS cathode load diff amps. i don't give them any NFB inputI just realized what you're saying.
You are correct on one point. Without feedback, a differential stage with CCS on the cathode is perfectly balanced.
Your mistake is to assume that the balance is maintained when you close the feedback loop.
Measure with feedback... Imperfection redefined. Been there, done that!A single diff-stage just won't cut it! Try adding loop gain, but then you have to accept lots of NFB. Not that ideal!
You're not alone here... Some commercial "upgrades" assume the EXACT SAME FAULTY ASSUMPTION OF PERFECT BALANCE...
from you> > > Your mistake is to assume that the balance is maintained when you close the feedback loop.what sort of feedback loop are you talking about? I guarantee, I have not been using it. My NFB loop when needed is balanced and does not involve teh grid circuit of the diff amp.
regards,
Douglas
!
Beware the man who behaves deceitfully towards others, for there is nothing else to save you from his deceiving you
> My NFB loop when needed is balancedI'm not a fan of balanced circuits in general and balanced feedback is even more complex. Haven't tried it.
Theoretical analysis of a differential feedback amplifier shows sensitivity to variations in OPEN-LOOP GAIN in the multiple paths.
Only when that gain goes to infinity can we claim absolute balance and only in ideal zero-delay conditions. This inevitably means high level of NFB, which I try to avoid (based on my sonic evaluations).
I'd love to hear about your balanced solution if it's simpler and achieves a well-balanced drive to the final stage.
well, since both infinite gain and zero delay are not attributes of the system in question, the amount of NFB need not follow your prediction either. Of course it can...but, it does not need to in what I have come up with.
regards,
Douglas
!
Beware the man who behaves deceitfully towards others, for there is nothing else to save you from his deceiving you
> I found no measurable un-balance with CCS cathode load diff ampsThat's very interesting, because I SPICEd this so meticulously and even the theory says no go. It was my first try, just that it didn't do the job and I do not want to risk saturating the PP OPT.
I'd consider interstage tranny to correct that imbalance, but that's a whole new story.> Split load inverter is only going to have a maximum output of 25%
That's in the ballpark. Triode-strapped 6550 only needs 50V peak-to-peak for full power, easily achievable with 480V supply. I myself had doubts about the direct connection, until I realized that the unity-gain splitter circuit is based on high local feedback, which stabilizes also the DC operating point. As long as you bring the tube to a decent operating point by design, it will adjust (based on tolerances) to still operate close to that point. The major sensitivity is to variations in plate voltage of the first stage.
In fact, the no-signal bias of the splitter is around -2V. The grid and cathode "move together" in the presence of signal, thus maintaining the bias. An extremely safe design!> The NFB-less triode V-A and direct coupled split-load did not ever measure well
My estimate is less than 2% THD at 10 watts (the amp is capable of much more power), mostly coming from the output stage. As I said, many dual-triode tubes would run this amp to a few watts output without much of a problem, but few can avoid clipping at large dynamic swings (close to full power). You must be careful what you measure. This is a DIY project, so you better "qualify" a tube for the application. I found mid-60's Sylvania 12AT7 very good and my friend tried United Electronics 801's successfully. Some tubes sound horrible, apparently not meeting their large signal specs.
I should have used the distortion analyzer while I still had the amp here. Sine wave was not visibly squished and crossover was smooth. It didn't sound distorted, but the same can be said for an SE 2A3, which measures miserably...
In any case, it sounded very fast and produced realistic tone and good dynamics. Bass was tight and lively and cymbals sounded totaly clear and undistorted to my ears. I've heard recently so many outrageouly expensive tube amps that won't hold a candle to this modified MkIII. Naturally you won't use it to power Wilson Sophias in an auditorium, but it would be extremely musical and tuneful on Sonus Faber Amatis in a large living room, for example.
And you won't feel like the Amatis deserve a better amp!
we're talking about 2 different things, you, the whole amp ( I think ) and I am only refering to the front end, loaded a little bit lighter than it would be in service. plus a 1M/40 pF probe circuit, alos balanced. They all turned square waves to something like a saw tooth by 10k, not to mention being imbalanced.Add a loop NFB and it can get fairly clean at the amp's output.
regards.
Douglas
!
Beware the man who behaves deceitfully towards others, for there is nothing else to save you from his deceiving you
I wouldn't get too stuck on square wave response, despite the popularity in the "free press". My amp has zero feedback, so the front-end response is just that - and it shows a tremendous rise fall time.
The front-end by itself extends to well over 500 KHz, while the whole amp from input to 8 ohm output (I measured at 10 watts) extends to 80 KHz - without any global feedback.
The front-end produces a near-perfect 10 KHz square wave. The amp as a whole shows some smoothing of rise and fall, as one might expect. But do you really care when it extends beyond the last RIAA mandated pole (that very few implement anyway...), not to mention the roloff of your super tweeters at 40 KHz (typical) and your average ear at 13 KHz (as I found in a group test!)...
quoted> > > The front-end produces a near-perfect 10 KHz square wave.into what load? try it through equal coupling caps and a grid resistor. Add to this a scope probe and it will probably look like a power valve grid, or at least enough to make some judgements.
Don't worry, not hung up on the square wave response, just a measurement ref. and test method. WIth loads other than the 'scope probe, response sucked on my test rig. B+ in the 420V range, plate V of the amplifier stage about a fourth...
Clearly we are doing something different, else our results should match more closely.
regards,
Douglas
!
Beware the man who behaves deceitfully towards others, for there is nothing else to save you from his deceiving you
> try it through equal coupling caps and a grid resistor.Load was 100K bias resistor and a standard probe.
The difference is a 12AT7 at 4 mA vs. a 12AX7...My friend built a second amp and measured 175 VDC on the voltage amp plate and 130 VDC plate-to-cathode on the splitter, just intended by design.
Start with a 500 volt supply, a 12BY7 for the voltage amp, and direct couple it to a 7233 running with 6k loads...I'll bet it will run to past a hundred kHz before it thinks about dropping a volt out of 40.
regards,
Douglas
!
Beware the man who behaves deceitfully towards others, for there is nothing else to save you from his deceiving you
This post is made possible by the generous support of people like you and our sponsors: