The State Variable oscillator looks like it should work pretty well. However the diode limiter for "agc" will add a lot of distortion for the system to remove. I don't understand the extra signal from the 91K resistor. Does it do some harmonic cancellation? If you use a linear AGC (even the thermistor from the Wein bridge circuit would be good) that circuit should get to .005% THD with TL072's. With higher performance opamps you can get to .0005% or less easily. The non-inverting output amp will have higher distortion due to the common mode modulation on the input circuit. But it will also have much lower noise. If you use opamps with some poop- even 5532's then ditch the output follower and go straight to a level control.
If fiddling with oscillators is not exciting get one of Victors oscillators- SOTA THD and cheap http://www.diyaudio.com/forums/equi...on-audio-range-oscillator-21.html#post3116055 You will be hard pressed to measure its performance with the best commercial equipment.
If fiddling with oscillators is not exciting get one of Victors oscillators- SOTA THD and cheap http://www.diyaudio.com/forums/equi...on-audio-range-oscillator-21.html#post3116055 You will be hard pressed to measure its performance with the best commercial equipment.
That's a nice design and shows clear understanding of the distortion mechanisms. I've done similar buffering of the drain voltage to do the half-drive to the gate, and it helps a lot. In addition he's got the voltage swing minimized to begin with, with the JFET close to virtual ground. A still-longer-channel JFET would probably enhance performance even more, but the part shown is readily available.
I don't have the patent handy, in fact I've only heard about it, but Hofer's innovations for amplitude stabilization of an oscillator, probably done while he was still at Tektronix, allows rapid settling while changing the frequency without a lot of control loop feedthrough. It may be one of those patents with an obscure name, but ought to show up in a search using Bruce Hofer.
This is the distortion null on the simple SVF ( A 10K 91K B ) . 51K in series with the 1N4148's and 22K across them . Note how it is not a square wave . I designed this for 50 or 60 Hz . Also 40 to 125 Hz ( Garrard 501 ) . 3 kHz seems to be a limit without some tweaking . I am using 220 pF NPO and 240 K 50 ppm . same results using 4n7 . The 220 pF was to check if low resistance was the problem at high frequency . The problem is the output rises and with it distortion . Could be 33K + 1 K needs trimming with frequency . I will check . Posting this as many designs far better than this use S V F . A tad better than I claimed ( 0.08 % ) . However a nice turntable oscillator with useful 3 kHz . There is no point going for better on a turntable as the distortion of the rotor is higher ( 1.5 % typical , 10 % synchronous ) . One keeps it sensibly low .
Off on holiday , look out Rhodes . Made a resolution to build something better when I get back .
I used a similar oscillator for a high end turntable (Rockport) that had to be continuously tuneable from about 15 RPM to 85 RPM. Getting two phases at 90 degrees makes the drive electronics much simpler. As you say reducing the harmonics on the voltage may do nothing useful. I also explored current drive but could not be sure if that helped either. It did make varying the drive with frequency easier. Fortunately the mass of the platter was high enough (45 lbs. of Stainless) to hide whatever nasties the motor produced.
Off on holiday is 2 hours , 5 AM on Sunday but leaving to stay with a friend near the airport . Couldn't leave without sorting the SVF . This is a good use of the spare op amp . It both sets gain and helps do a clean up . Has worked on strip board .
220 pF was to prevent any possibility of low impedance problems . Chip used TL074 . I suspect 1nF a better choice . - 70 dB 0.032 % 3kHz ( 0.04 % THD ) . TL074 will work well with calculated values . It also needs no bias current to work ( little ) . Output varies a little . Generally 6 V rms ( about +/- 1 dB ) . Useful for testing buffers . I dare say it will go higher in frequency .
Useful cheap tool for the workshop and great for measuring tube amps . +/ - 15 VDC .
This shows how to get cosine and sine . Extra filtering if wanting low distortion .
http://www.massmind.org/images/www/hobby_elec/e_ckt20_2.htm
Last edited:
Had to check 5532 against TL072 in the Wien Bridge ( Hewlett Packard invention I believe ) with RA 53 . No noise advantage noted either way ( 68 K not ideal into op amp 2 ) . The Wien has the slight advantage over SVF if you can find a RA 53 or workable torch bulb ( other feedback arm if so ) . 5532 has output advantage so will stay . Humble LM358 ( 324 ) was - 66 dB about like simple SVF ! Note how op amp changes frequency .
Patent US4560958 - State variable oscillator having improved rejection of leveler-induced ... - Google Patents (The Google patent interface is way better than the USPTO)
What you do is prime the tuning capacitors that are not in use. Charge them with a dc voltage equal to the rms of the operating level. The caps are charged in a quadrature relationship. When the caps are switched in the oscillator is in a ready to go state. This minimizes the settling time during range changes. This was done in the SG505. You can see the arrangement in the schematic at the top of the range switches.
Last edited:
All of this brings us back to the fundamental questions:
1. Exactly how much THD is needed to become audible, even if it's not the actual levels but the distribution of the harmonic decay, and
2. Consequently, how low do we need our THD measuring equipment to be reliable to - 0.1%, 0.01%, 0.001%, etc?
My personal view is that as long as overall THD levels, operating into actual loudspeakers rather than lab resistors (or a more complex and realistic model thereof - see below) is kept to 0.1% or less, at nominal power output into the least rated nominal impedance (assumed to be say 4 Ohms nominal, minimum say 3 Ohms), then that's all we need. Observe the THD levels of a typical zero global NFB tube amp as an explanation - sometimes looking criminal, at 3% or more, but nevertheless sounding just great.
I would use three versions of this test - at nominal power output, at half and at one quarter nominal power. In other words, full rated power is "reserved" for short time peaks only, as I assume no-one here wants to hear what's clipping like and no-one is placing bets on the lifetime of the speaker drivers under clipping conditions.
Personally, I use a model of Infinity's Reference 5 (later 50) speakers. It's a 3 way system which I happen to have some experience with and know that while not really BAD are still no joke of a load to drive, and quite a few commercial, low on the scale models will trip on when driven hard. "Hard" meaning 1/4 of nominal power.
Lastly, I find that increased THD levels are incomparably preferable to other possible sins inside a power power amp, such as, for example, current starvation for any given mode of operation. So the tympani sounds like a guy bullying a helpless pot.
1. Exactly how much THD is needed to become audible, even if it's not the actual levels but the distribution of the harmonic decay, and
2. Consequently, how low do we need our THD measuring equipment to be reliable to - 0.1%, 0.01%, 0.001%, etc?
My personal view is that as long as overall THD levels, operating into actual loudspeakers rather than lab resistors (or a more complex and realistic model thereof - see below) is kept to 0.1% or less, at nominal power output into the least rated nominal impedance (assumed to be say 4 Ohms nominal, minimum say 3 Ohms), then that's all we need. Observe the THD levels of a typical zero global NFB tube amp as an explanation - sometimes looking criminal, at 3% or more, but nevertheless sounding just great.
I would use three versions of this test - at nominal power output, at half and at one quarter nominal power. In other words, full rated power is "reserved" for short time peaks only, as I assume no-one here wants to hear what's clipping like and no-one is placing bets on the lifetime of the speaker drivers under clipping conditions.
Personally, I use a model of Infinity's Reference 5 (later 50) speakers. It's a 3 way system which I happen to have some experience with and know that while not really BAD are still no joke of a load to drive, and quite a few commercial, low on the scale models will trip on when driven hard. "Hard" meaning 1/4 of nominal power.
Lastly, I find that increased THD levels are incomparably preferable to other possible sins inside a power power amp, such as, for example, current starvation for any given mode of operation. So the tympani sounds like a guy bullying a helpless pot.
Last edited:
I would add that THD and all steady-state performance is insufficient, and possibly less important than transient response.
Besides that, why limit THD testing to audio range? Why not increase the frequency until we find out where the drop-off occurs?
P.S. I concur on current starvation. But that's easy if you design as if for constant DC output capability, at the peak (not RMS) sine level implied by the rated max power. You could test it by lowering sine input frequency while testing THD, at max rated power. But you could instead DC couple and test with DC at the max peak level, into a dummy load of course.
Besides that, why limit THD testing to audio range? Why not increase the frequency until we find out where the drop-off occurs?
P.S. I concur on current starvation. But that's easy if you design as if for constant DC output capability, at the peak (not RMS) sine level implied by the rated max power. You could test it by lowering sine input frequency while testing THD, at max rated power. But you could instead DC couple and test with DC at the max peak level, into a dummy load of course.
Last edited:
Both necessary, IMO. Transient response appears more important because steady state, and noise, were done and dusted some time ago.
I'm still intrigued by Pano's cello. I am working on the hypothesis that, "sounds like a real cello" is a superposition of "sounds like a cello" and "sounds real". The real bit corresponds to whatever people mean by "natural".
It's tempting then to attribute the "sounds like" to the steady state response, and the "sounds real" to the transient response.
It then seems to follow that the system responsible for Pano's cello, at a hi-fi exhibition some time ago IIRC, happened to be in a spot that suited it, whereas some with "better" specifications did not.
I come from the direction of control systems rather than electronics or audio, and I'm still bemused by some EE conventions. In particular, the use of the word "transient" as a noun really makes me cringe, even though I accept that, linguistically, usage trumps logic.
I hate it because that usage tends to confuse the very meaning of the term "transient response", to the extent that if I ask "what is the ideal transient response of an audio system", nobody seems to know the answer.
I like the look of waterfall charts...they seem to point in the right direction, but I find them hard to interpret. I wonder if there is a way to reduce one to a brief set of figures of merit.
I wish I'd got into speaker design when my mind worked faster. That's where the good stuff is. Whole-system transient response seems to be in the horribly complicated stage of analytical development.
I wish there were more people here who understand control systems. Then we would get less nonsense about feedback.
I learnt about NFB from electronics books. Then I learnt about servo systems from some OU notes when I worked for the power industry. Then I realised that they were the same thing.
I learnt about NFB from electronics books. Then I learnt about servo systems from some OU notes when I worked for the power industry. Then I realised that they were the same thing.
One of the sharpest guys I know studied control systems as part of his mechanical engineering curriculum. He started at Harman as mostly a speaker designer, and out of his frustration with some EE people wound up teaching himself about electronics. I mentioned his prowess to a friend there, how clever he was, and the friend admitted to being barely able to keep up with him
@gootee
I never said audio range only. Hardly could, as I am concerned what it does at 50 kHz, both with no global NFB and with it.
Think back and you may remember that I repeated several times I like my amps to reach above 300 kHz at full blast, and then like reVox/Studer and Sony, to limit them by inserting a low pass filter at around 200 kHz.
Unlike some folks here, I still abide by Otala's ideas of a wide open loop full power response, do my best to reduce distortion as well as I can, and then add just enough global NFB to keep the lot stable. This typically means 20 to 26 dB of global NFB, but is not an iron clad rule.
More often than not, I find getting 40 kHz of open loop bandwidth is well within my reach, on occasion even more, with lower powered amps (say 50/100W into 8/4 Ohms) even as high up as 80 kHz. When your HF response extends to 80 kHz, I find you don't really need all that much of global NFB anyway for very reasonable results, even for transient (sorry, DF96) good behaviour.
But that assumes generous power supplies. The said 50/100W amp uses 400 VA toroids, one for rach channel, followed by 2 x 10,000 uF of caps per line, or 80,000 uF overall. With local (i.e. right next to the output transistor) additional caps, rated at 2,200 uF each. And with separate. electronically regulated input stage and VAS power supplies.
As you can gather, I really, but REALLY want my tympani to sound like tympani, and looked at it from that angle, I am not overdoing it at all. If you want current, it has to come from somewhere. So I use two pairs of Motorola/ON Semiconductor 200W devices per channel, and I spend time pairing them up from the bulk I own.
I am probably going a little overboard, but if I am to err, I want to err towards the overboard side.
I never said audio range only. Hardly could, as I am concerned what it does at 50 kHz, both with no global NFB and with it.
Think back and you may remember that I repeated several times I like my amps to reach above 300 kHz at full blast, and then like reVox/Studer and Sony, to limit them by inserting a low pass filter at around 200 kHz.
Unlike some folks here, I still abide by Otala's ideas of a wide open loop full power response, do my best to reduce distortion as well as I can, and then add just enough global NFB to keep the lot stable. This typically means 20 to 26 dB of global NFB, but is not an iron clad rule.
More often than not, I find getting 40 kHz of open loop bandwidth is well within my reach, on occasion even more, with lower powered amps (say 50/100W into 8/4 Ohms) even as high up as 80 kHz. When your HF response extends to 80 kHz, I find you don't really need all that much of global NFB anyway for very reasonable results, even for transient (sorry, DF96) good behaviour.
But that assumes generous power supplies. The said 50/100W amp uses 400 VA toroids, one for rach channel, followed by 2 x 10,000 uF of caps per line, or 80,000 uF overall. With local (i.e. right next to the output transistor) additional caps, rated at 2,200 uF each. And with separate. electronically regulated input stage and VAS power supplies.
As you can gather, I really, but REALLY want my tympani to sound like tympani, and looked at it from that angle, I am not overdoing it at all. If you want current, it has to come from somewhere. So I use two pairs of Motorola/ON Semiconductor 200W devices per channel, and I spend time pairing them up from the bulk I own.
I am probably going a little overboard, but if I am to err, I want to err towards the overboard side.
One of the sharpest guys I know studied control systems as part of his mechanical engineering curriculum. He started at Harman as mostly a speaker designer, and out of his frustration with some EE people wound up teaching himself about electronics. I mentioned his prowess to a friend there, how clever he was, and the friend admitted to being barely able to keep up with him
That's the thing, Brad. Being self-taught has one downside and one upside above all else:
The downside is that you are often missing a bit of theory here and there, so you just as often need to go clean up.
The upside is that that you are not saddled with anything, you are much more free to think outside the box of the narrow profession, and it sometimes pays off big time.
This guy to whom I refer has gone on to a startup and is doing the electronics. He emails me from time to time with an issue for which, understandably, he doesn't yet have the background. I'm hoping that sooner or later the startup is sufficiently successful that I can do some billing
But then most engineers emerging from a four-year school don't know anywhere near a truly comprehensive body of knowledge --- nor should this be surprising, given the incredible breadth and depth of the field. I think I mentioned before, that one seasoned professional said he'd only consider hiring someone fresh out of school if she or he had studied with one of only two (!) professors, either Meyer at UC Berkeley or Blesser at MIT. Neither is still teaching now.
One bright guy I know who also worked at Harman for a while and had a degree from Cornell once asked me, somewhat sheepishly, how a switch worked. It took me a while to understand his question. He literally did not know, nor had ever been taught, that inside, say, a miniature toggle, there were metal pieces that were brought into physical contact or pulled apart. Such can be the gaping holes in one's experience and understanding.
Just this morning, I had to tell my 27 year old son that since I am no longer 27 years old, I do NOT know everything. And I harbour some doubts.
The idea of hiring kids straight out of college is not at all a bad idea. Otherwise, they might be "taught" wrong, so you'd first have to "unteach" them the wrong and then to teach them the right way. A hell of a job - good luck to anyone who has to do it.
For the fun of it, try talking to a fresh college graduate about ground - most think that ground is a ground is a ground, and the mere concept that this just might not be quite so simple baffles them. I stumbled into one such discussion some years ago. Explaining this concept was harder than giving birth to a baby bear.
The idea of hiring kids straight out of college is not at all a bad idea. Otherwise, they might be "taught" wrong, so you'd first have to "unteach" them the wrong and then to teach them the right way. A hell of a job - good luck to anyone who has to do it.
For the fun of it, try talking to a fresh college graduate about ground - most think that ground is a ground is a ground, and the mere concept that this just might not be quite so simple baffles them. I stumbled into one such discussion some years ago. Explaining this concept was harder than giving birth to a baby bear.
- Status
- Not open for further replies.