Thanks,
I'd just located a copy of your board photo I'd stored. I hadn't been on DiyAudio for some time. I'm still having issues navigating posts and threads.
Hi Perry,
I hadn't thought much about simulation but, silly me, that's what I usually do.
I've finished and attached my Rev 01 driver schematic, as both an LTSpice .asc file and as a pdf, using the visible designators on my black board. I haven't yet done your blue board. I was using LTSpice strictly for schematic capture. To simulate, I'll have to do a wee bit more work. While the circuits look semi-reasonable, I haven't gone through everything in detail. A simulation will quickly disclose any goof-ups.
If you use LTSpice, their digital AND and OR gates are all 5 input, so I've jumpered 4 inputs together to create each 2 input gate.
Measuring unmarked ceramic in-circuit capacitors is a pain. I've never asked (until this very moment) why aren't SMT capacitors value-marked like resistors? Here are some comments:
https://forum.allaboutcircuits.com/...be purchased with,so many forgoe the markings.
My True rms DMM in Capacitor Mode doesn't like capacitors shunted by resistors whose R is less than the capacitor's Xc at the test frequency
If you want to measure in-circuit capacitance accurately, you need to apply a sine wave Voltage and measure the in-phase Current for resistance and the quadrature (90 degree leading) Current for capacitance. Like a Wheatstone bridge. A jelly bean multivibrator cap tester doesn't work well.
BTW - if you've ever used a water purity tester, they're AC Ohmmeters. DC Ohmeters polarize the water and screw up the readings.
Fortunately, I have a small USB RLC board that does a credible job with in-circuit capacitance.
Please let me know what you think.
I was too busy to call Soundstream today. With your help and my own driver schematic, I don't really need them.
What may(?) have turned them off - I politely explained I'm a medical equipment designer, doing infrasonic studies. Didn't tell them I'm really an audio madman, bent on destroying my SUV and whatever hearing I have left, with 16 Hz classical pipe organ music, drums and thumping dinosaurs.
I'm apparently too dumb to be devious.
Thanks again for your help. Without it, I'd still be struggling.
My beefier connector pins and headers arrived. I've installed them, with the driver board tilted back, 45 degrees, over the analog front end stuff, since the header raises the driver board about 3/8 inch.
Ron
* If it takes 10 Watts at 64 Hz to produce a 100 dB SPL, 1 meter from a sealed box, it takes 160 Watts at 32 Hz and 2.5 kW at 16 Hz. I'm trying to duplicate my family room's Bob Carver Class H Sunfire True Subwoofer in my SUV. I could have simply run another Sunfire from a 12 VDC to 120 VAC inverter. That's expensive and no fun, but it's always an option.
The BK Precision 879B does a pretty good job measuring capacitors in the circuit.
The Carver sub could have been modified to use a custom switching power supply in place of a 12v-120v inverter.
The first email I send a manufacturer is one simply asking if the email is still a working email. Nothing else.
In the car, you have to remember that you will get a lot of help from the transfer function of the vehicle.
Unless I'm mistaken, the carver sub uses a very high excursion woofer and passive and to overcome the back EMF of the woofer, they simply rectify the full mains voltage to get 170v of rail voltage.
If you try the Arduino breadboard jumper wires, I think you'll find them to be much easier to work with than the rigid connectors, especially when removing them when the repair is done.
The Carver sub could have been modified to use a custom switching power supply in place of a 12v-120v inverter.
The first email I send a manufacturer is one simply asking if the email is still a working email. Nothing else.
In the car, you have to remember that you will get a lot of help from the transfer function of the vehicle.
Unless I'm mistaken, the carver sub uses a very high excursion woofer and passive and to overcome the back EMF of the woofer, they simply rectify the full mains voltage to get 170v of rail voltage.
If you try the Arduino breadboard jumper wires, I think you'll find them to be much easier to work with than the rigid connectors, especially when removing them when the repair is done.
The Carver sub could have indeed have been driven by a 12 VDC to 120 VDC switching supply. I wasn't dedicated enough to build one. It had to be cheap, off the shelf and somewhere in the USA. Plus. if I recall, the Carver doesn't have a soft-start MUTE. I'd have to do that, too.
What is the transfer function from the vehicle? Is that all the power switching stuff?
When I started the subwoofer project in my 6 month old 2016 RAV4, I was simply going to use the mfr's built-in Class D subwoofer amp - the problem - the audio path was digitally equalized to boost the RAV4's 35 Hz rolloff crappy 10 inch "subwoofer". That's what started this whole affair, 6 years ago. The factory sub and amplifier wouldn't do my desired 16 Hz or even Carver's 18 Hz. My 1st generation shallow 12 inch Infinity driver crapped out ~ 25 Hz. That's when I was reminded that shallow subs stink and when a mfr says "minimum usable frequency", that's often where the cone does the drumhead, eigenmode breakup.
The Infinity, mounted under the tonneau cover bar for safety and stability. Had to cover it with a cardboard panel whenever the grocery pickup people filled the rear deck. What you can't see is a 10 inch Earthquake drone, on the lower right backside, tuned to ~ 22 Hz. I don't especially care for the dynamics of ported enclosures, but it kinda almost worked.
But as I was told and subsequently realized the hard way, shallow cones are bad for bass. Try doing the shallow/deep cone stiffness experiment with a sheet of paper.
My second 12 inch subwoofer had a deeper Aluminum cone. I still don't trust Carver look-alike drivers. I couldn't find a 12 inch, dual VC sub with a sub-20 Hz specified breakup that wouldn't break the budget. I didn't want to buy a used Carver subwoofer. At the time, I also wanted the 2nd VC for motional FB. Somewhere, I'd mentioned using accelerometer. pressure gauge, optical and, since I'm an old radar guy, even tried a $10 X-band (10 GHz) Doppler radar module, with some Aluminum foil on the dust cover. The Doppler module is another story.
The Carver has and does everything you'd mentioned. When it first came out, I debated between it and an 18 inch, 1st generation servo'd Velodyne. I liked Carver's combo of high SPL at 20 Hz and concealable size. I would have preferred Carver using the accelerator feedback they'd originally advertised.
Our condo at the time was tiny and there wasn't room for the larger, servo'd Velodyne. I was impressed with Carver's surround, built like a radial tire. Before I bought our first SONY 47 inch plasma, Carver's huge magnet raised hell with our 32 inch Trinitron's covergence.
I commiserated with Bob Carver. He'd patented his Class H Tracking Downconverter amp at about the same time I patented my 30 kiloWatt, water-cooled Class G MRI Gradient amplifier. He shelved his product for years, but finally made it work. My project was s--t canned, but I was able to bootleg a new system with Copley Control's multi-kiloWatt Class D antenna array slew amplifier that Dick Burwen had designed.
Other than buying the 1st generation Carver, I was out of DIY audio for over 15 years. Too busy working and raising a family.
Carter has taken lots of good ideas to product fruition. Some purist "audiophiles" don't like them or him. Let's see them do better.
By the way, have you calculated the stored energy in those 170 VDC caps? After my warranty ran out I played with extra weight on the passive radiator (drone). Then, one day, as I was probing, I slipped and vaporized all the output NPNs and their drivers. The itemized repair list filled an 8.5 x 11 sheet. I've still had to defeat the failure-prone sound-driven turn-on but it's still working, 24/7, all those years later.
The setup I now have allows me to use those Arduino jumpers if needed.
------------------------------------------------------------
I fired up the simulation and found several of the tied-together U4 NOR gate inputs are floating. "Should" be easy to find and fix. The simulation may actually be making baby steps today.
Last edited:
The transfer function is the low-frequency boost that you get from the cabin of the vehicle. It can be 20dB at the lowest frequencies.
Carver didn't miss a trick with those subs.
MRI. That's some scary tech. You're the second person who I've 'met' who designed MRI machines.
The energy goes way up with such high voltage. That's one reason I suggest testing with much reduced rail voltage for troublesome amplifiers that keep blowing output transistors when repairing them.
The driver board you have has muting. All you need to do is to add a circuit that goes from high to low voltage to go from mute to audio.
Carver didn't miss a trick with those subs.
MRI. That's some scary tech. You're the second person who I've 'met' who designed MRI machines.
The energy goes way up with such high voltage. That's one reason I suggest testing with much reduced rail voltage for troublesome amplifiers that keep blowing output transistors when repairing them.
The driver board you have has muting. All you need to do is to add a circuit that goes from high to low voltage to go from mute to audio.
MUTE isn't a big issue. I've started with it grounded. So far, I've only found one LTSpice IV-compatible comparator, similar to the LM119, fast, with higher than +/- 12V rails. I could always do it with a fast op amp, driving a grounded emitter, open collector NPN. The Analog Device comparator I've found is the Rad-hardened RH111, so in case my simulation is bombarded with high energy particles, I'll be safe.
The reason I haven't upgraded to LTSpice VIII, is my use of lots of interconnected log/antilog amplifiers. LTSpice VIII has convergence issues IV didn't. Without knowing LT's op amp macromodels, troubleshooting their convergence issues is a crap shoot. What's helped is converting from even LT's simplest op amps to Voltage-Controlled Current Sources.
This is a Gain of 1 Million simulated op amp, with a single-pole, 10 Hz rolloff, a 10 MHz Gain Bandwidth product and an open-loop output impedance of 100 Ohms, using NFB to reduce its gain to -10. The TVSs bound the output at +/- 5V.
If you want to observe the simultated op amp's open-loop GBW, increase Rfb to > 100 Giga Ohms or simply remove it.
The open-loop DC Gain of 1 Million comes from the VCVS's DC transconductance of 10,000 Amps/Volt, times the open loop DC output resistance of 100 Ohms or 100 Volts per Amp.
(VCVS transconductance)(Rout1) = Voltage Gain
(10,000 Amps/Volt)(100 Volts/Amp) = 1 Million
all the units cancel.
We used to joke that the huge magnetic fields around MRI's superconducting magnets would do strange things to the minute amounts of lodestone or magnetite in our brains and make us silly. Research now shows that very same lodestone is used by common pigeons for navigation. They can sense minute distortions of the Earth's magnetic field around steel fences, steel downspouts, gutters, car bodies, etc.
About 25 years ago, NIH Bethesda installed 2 giant, 5 Tesla MRIs and they were too cheap to buy LCD monitors. The staff within several hundred feet were still using color CRTs inside half-assed mu-metal shield boxes. Screen images and colors were all buggered and tilted.
Thanks for reminding me.
I wasn't thinking of cabin boost. That's what I was trying years ago to get from a ~ 100 dB SPL to ~ 120 dB.
I found I'd mislabeled and shorted +12F (+12V Filtered) to -12F and and couple of +5 points that were Text-labeled +5, weren't actually connected to the +5 bus.
The modulator works fairly well. The triangle looks good with a 100 Hz sine wave input at the modulator input. So far, I haven't measured modulator linearity or done any sort of power stages or overall analog NFB.
Hi Perry,
I found a few more small errors. The attached Rev 02 Simulation works nicely. The triangle is ~ +/- 1.1 V at 110 kHz and visually quite linear. The PWM Modulator can handle ~ a +/- 0.9 V sine wave.
The two RC/diode networks create non-overlapping square waves to the input LEDs of the two digital optos whose outputs drive the pre-output FET emitter followers.
The overall FEEDBACK loop (not yet connected) should reduce non-clipping rms THD to "well below" 1%. I'll try it and measure THD, when I simulate the entire output power stage.
I've been unable to find MOSL2 and MOSH2 on the main board schematic. The two parallel 2 Ohm resistors provide a GND reference point for Voltage or Current probing. I'm not losing my mind (yet). The X10 V/V is for amplifying the PWM's audio input for simultaneous viewing of it and the 5V digital signals.
Ron
Last edited:
Hi Perry,
I should have realized the two diode/RC networks, intended to prevent "punch-through", the simultaneous conduction of the upper "N" FET bank and the lower "P" bank, create an onerous crossover dead band. At low levels, without overall FB, the Soundstream is a Class C amplifier.
If you short the crossover bias diodes or VBE multiplier(s) of a classical bipolar transistor Class AB amplifier, there's a small gap or level area in the output transfer characteristic, typically 2 VBEs wide. It's even worse if the output devices are Depletion mode FETs. Almost all power FETs are Depletion mode, not Enhancement mode.
A timing gap in a Class D produces the same distortion as a Voltage gap in a linear amplifier.
An FFT of my simulated amplifier, at 100 Hz, driving 500 Watts into 2 Ohms, without overall FB is ~ 7%. I haven't tried it yet, but a low output levels, THD without FB could approach 100%.
I was able to apply 36 dB of overall FB, which reduces the 500 Watt THD to ~ .09%, but taking FB before the output LC mixes a small triangle with the comparator's analog audio input. I'm not quite sure what that does in the time/frequency domain or to the audible output noise.
It's a classical designer's dilemma - a good design fixes the real problem without using feedback or another patch to reduce it. Since I've never been inside a Class D audio power device, I need to look into solutions before convincing myself that FB works well enough.
Advanced Class D devices use things like Delta Sigma modulation/demodulation. Some suggest something simpler, like dithering the triangle generator or opposing, non-overlapping, narrow output pulses.
"How do I fix Class D Crossover Distortion?" - the question and solutions go beyond automotive amps - I need to look at more Class D Crossover Distortion threads. My results, so far, haven't been encouraging. I may have to ask a moderator for advice.
My Best,
Ron
Last edited:
Most class D amps use feedback either before or after the output filter.
The infinitely small deadtime of the output FET switching prevents distortion due to switching. It's a balance where shoot-through is right on the margins.
I don't think I've ever seen crossover distortion in a class D amp. The FETs are fully on as they cross 0v.
The infinitely small deadtime of the output FET switching prevents distortion due to switching. It's a balance where shoot-through is right on the margins.
I don't think I've ever seen crossover distortion in a class D amp. The FETs are fully on as they cross 0v.
For the first time in 2 years, I decided to go to bed before midnight. So what happens? I'm fully awake at 4 AM, wondering how to fix Class D crossover distortion. My wife still asks me "What's wrong?"
The SS designers had to make the timing gap wide enough to prevent significant shoot-through v supply voltage, temperature, etc. If nothing else - it gooses the supply lines, the low level analog stuff and nearby electronics.
That brings up the question of what SS and others do in their full bandwidth Class Ds. I may be forced to find one, or at least a schematic.
Possibilities:
1) use SS's FB to add triangles to the analog audio. It's simple and works. I'll FFT it to see how well.
2) find a feedback way to make the output devices actively and reliably sit on the very edge of shoot-through. Years ago, National Semi had an intelligent AB crossover bias chip. I'm probably the only one who ever even read the data sheet, much less even considered using one.
3) dither the PWM converter. I actually don't know what that means or entails. I've simulated a number of Linear Feedback Shift Registers (LFSRs), but wasn't even smart enough to use one to flicker the yellow LEDs in an electronic candle.
4) mod the output stage so an upper/lower pair or all of them behave like a classical, linear AB at low signal levels
5) combine a classic Class AB, linear output stage with the Class D - that's essentially the same as Possibility 4).
6) use Delta Sigma stuff? Am I really that crazy? I'll still look at data sheets and publications before I cave and do 1).
Remember, who was it that said "If you can't define the math, you don't understand the problem"? I like "If you can't at least simulate it, you have no hope of understanding it.".
I'm not sure if it's this fellow who haunts my messy lab.
Before the Soundstream, it was a lot neater. It's bad enough I'd volunteered to fix my neighbor's connector-ridden, multi-function SONY. The SOBs buried the power supply in a literal cave, under the PCB.
I'm not inclined to build anything as beefy as an amplifier - for the first time in my life, I actually bought an electronic pegboard, because these are my typical audio breadboards. You should see the others.
Last edited:
I've done plenty of the above: notebook scribbles/modeling/simulations. I don't go anywhere without my ubiquitous 8.5 x 11 quadruled spiral notebook. Everything from a full sized calendar to math, schematic scribbles, to the grocery list. Too much for my tablets or phone. I fill a notebook every 4 months. Every few years, I scan then shred about 10.
------------------------------------------------------------------
One other thing - my Soundstream has what looks like a +/- 60 V supplies for the push-pull output FETs. I haven't actually measured the rail voltages. The rails can't be any higher, since the cap on each rail is rated at 63 Volts. That's already too close.
The output is a straight push-pull. It isn't a full H-bridge. One side of the speaker is grounded.
Assume the output load is 2 Ohms, resistive. The amplifier is rated at 4.5 kW, burst mode, into 1 Ohm, which equates to half that or ~ 2.25 kW into 2 Ohms, resistive.
Assume a 5% supply dip at full output power, including IR drop in the output inductor. Let's further assume the regulator loop isn't fast enough to limit supply dip for a short 100 Hz burst.
The max unclipped sinewave output is ~ +/- 57 Volts or 114 V p-p. Dividing that by 2[sqrt(2)] or 2.828 gives a 40.3Vrms output sinewave.
(43.2Vrms squared)/(2 Ohms) = 933 Watts, burst mode, into 2 Ohms or 1.87kW into 1 Ohm, resistive.
Now, if it were a full bridge, you'd expect ~ 4X that or 3.7 kW, burst mode, into 2 Ohms. That's 7.4 kW into 1 Ohm, resistive.
Am I missing something here?
Maybe the output supplies are really ~ +/- 95 Volts under burst load? That means the two 63 Volt caps in my amp photo are in series for one rail, but then, where are the other two series caps for the other rail?
------------------------------------------------------------------
One other thing - my Soundstream has what looks like a +/- 60 V supplies for the push-pull output FETs. I haven't actually measured the rail voltages. The rails can't be any higher, since the cap on each rail is rated at 63 Volts. That's already too close.
The output is a straight push-pull. It isn't a full H-bridge. One side of the speaker is grounded.
Assume the output load is 2 Ohms, resistive. The amplifier is rated at 4.5 kW, burst mode, into 1 Ohm, which equates to half that or ~ 2.25 kW into 2 Ohms, resistive.
Assume a 5% supply dip at full output power, including IR drop in the output inductor. Let's further assume the regulator loop isn't fast enough to limit supply dip for a short 100 Hz burst.
The max unclipped sinewave output is ~ +/- 57 Volts or 114 V p-p. Dividing that by 2[sqrt(2)] or 2.828 gives a 40.3Vrms output sinewave.
(43.2Vrms squared)/(2 Ohms) = 933 Watts, burst mode, into 2 Ohms or 1.87kW into 1 Ohm, resistive.
Now, if it were a full bridge, you'd expect ~ 4X that or 3.7 kW, burst mode, into 2 Ohms. That's 7.4 kW into 1 Ohm, resistive.
Am I missing something here?
Maybe the output supplies are really ~ +/- 95 Volts under burst load? That means the two 63 Volt caps in my amp photo are in series for one rail, but then, where are the other two series caps for the other rail?
The JL 1000 is the only amp that I've ever seen with series-connected rail caps.
I'm not sure what regulation you're referring to but if this amp is like the sla1500, there is no regulation for the rail voltage. In the real world with voltage drop in the B+ supply line and loss in the switching supply, I'd expect more than a 5% loss but that's with a normal amp. I don't know what you're going to be using for a power supply.
If the rail caps are 63v, that's the limit for the rails (±63v max).
I'm not sure what regulation you're referring to but if this amp is like the sla1500, there is no regulation for the rail voltage. In the real world with voltage drop in the B+ supply line and loss in the switching supply, I'd expect more than a 5% loss but that's with a normal amp. I don't know what you're going to be using for a power supply.
If the rail caps are 63v, that's the limit for the rails (±63v max).
The TL494 is designed to be both a Voltage and Current PWM regulator. Whenever pin 1 is more positive than pin 2, op amp 1's output swings positive. Whenever pin 16 is more positive than pin 15, op amp 2's output swings positive. Each output is connected through a diode to the chip's feedback pin 3. The pin 3 line also drives a gate that shortens the output pulse. Whichever of the two output pins is more positive, it dominates the pulse shortening process. If either regulated Voltage or Current passes a preset threshold, the reduced pulse width decreases the power circuit's Voltage and/or Current.
In most modern power supplies, either the critical or the highest Voltage/Current is/are regulated and, because of the transformer turns ratios, the other Voltages nominally track. But they sometimes have lousy load regulation.
All that said, I have yet to find find any direct evidence SS's 494 regulates any particular voltage.
-------------------------------------------------------------------------
The supply rails - if the amp uses 1 cap per rail , then the max rail voltages are + and - 60 V, unloaded, with like a 14.5V charging battery. That's pushing the caps. If the unregulated rails droop 10% under a short burst output load, plus 2% IR droop in the conducting FET bank and another 2% IR drop in the output inductor and PCB, the rails are effectively +/- 51.4 Volts.
1) I'm re-reading the SS AR1.4500D Box AND the Amp's data sheet. The box says 4,500 Watts and the data sheet says 2,200 Watts rms, both into 1 Ohm at < 1% THD, which is essentially unclipped. I assume the 4,500 Watts is burst mode. The best +/- 51.4 V rails will provide is
[( [(102.8V p-p) / (2.828)])^2]/ (1 Ohm) = 1,321 Watts, nowhere near the box's 4,500 Watts and still well below the data sheet's 2,200 rms Watts.
2) While everyone uses it, there is no such thing as an rms Watt. If you multiply in-phase rms sinewave Voltage times in-phase rms sinewave Current, you get Watts, not rms Watts. If they're not in phase, you get Volt Amps:
Volt Amps = (rms Volts)(rms Amps)(cosine of the angle between them)
or, if they're distorted, you have to root sum square each and every separate harmonic, including the cosines. That's why properly done simulations and FFTs are great.
What they should be saying is Sustained Watts. They're all trying to BS people. They might as well be speaking Latin or, better yet, the WW2 Code Talker Navajo language.
For the task I'm doing < 1% THD is important, while 650 Watts, burst mode, into 2 Ohms is fine for now. We can consider power as an academic exercise, since we both know people fib. I try not to test my tires over 70 mph and then only in Burst, not Sustained mode.
In most modern power supplies, either the critical or the highest Voltage/Current is/are regulated and, because of the transformer turns ratios, the other Voltages nominally track. But they sometimes have lousy load regulation.
All that said, I have yet to find find any direct evidence SS's 494 regulates any particular voltage.
-------------------------------------------------------------------------
The supply rails - if the amp uses 1 cap per rail , then the max rail voltages are + and - 60 V, unloaded, with like a 14.5V charging battery. That's pushing the caps. If the unregulated rails droop 10% under a short burst output load, plus 2% IR droop in the conducting FET bank and another 2% IR drop in the output inductor and PCB, the rails are effectively +/- 51.4 Volts.
1) I'm re-reading the SS AR1.4500D Box AND the Amp's data sheet. The box says 4,500 Watts and the data sheet says 2,200 Watts rms, both into 1 Ohm at < 1% THD, which is essentially unclipped. I assume the 4,500 Watts is burst mode. The best +/- 51.4 V rails will provide is
[( [(102.8V p-p) / (2.828)])^2]/ (1 Ohm) = 1,321 Watts, nowhere near the box's 4,500 Watts and still well below the data sheet's 2,200 rms Watts.
2) While everyone uses it, there is no such thing as an rms Watt. If you multiply in-phase rms sinewave Voltage times in-phase rms sinewave Current, you get Watts, not rms Watts. If they're not in phase, you get Volt Amps:
Volt Amps = (rms Volts)(rms Amps)(cosine of the angle between them)
or, if they're distorted, you have to root sum square each and every separate harmonic, including the cosines. That's why properly done simulations and FFTs are great.
What they should be saying is Sustained Watts. They're all trying to BS people. They might as well be speaking Latin or, better yet, the WW2 Code Talker Navajo language.
For the task I'm doing < 1% THD is important, while 650 Watts, burst mode, into 2 Ohms is fine for now. We can consider power as an academic exercise, since we both know people fib. I try not to test my tires over 70 mph and then only in Burst, not Sustained mode.
Last edited: