Wierd clipping & barking at slightly low line (110VAC)

[nevetslab]

I have a client's amp in the shop at the moment that's used to drive drum stool shaker motors. While checking it out, I found very peculiar barking sounds coming out of my 15" 4 ohm test speaker (not having a drum stool to drive, while sitting on it), and this behavior only begins happening at around 110VAC...even at 114VAC. @ 120VAC, solid as a rock, and clipping is symmetrical, without the nastiness.



The test signal is 50Hz Sine-Random noise, 10Hz BW modulation from a Bruel & Kjaer 1027 Sine Random Generator, which in this mode yields variable-amplitude LF sine wave. The slope on the peak waveforms where it is clipping suggests not enough power supply capacitance...though I don't think that's the cause of the behavior @ 110VAC driving to max. The schematic I have is too blurry to make out component values, and I haven't yet obtained a clean copy from the designer over in the UK. I sent him the scope shots to see if he's run into this before in production.

I haven't yet gone back to the amp to fill in the missing information (resistor values, xstr types, voltages at all the pertinent junctions. The nominal supply voltages are +/- 40VDC. Looks like part of the circuit is on the threshold of being starved, and with just slightly less than nominal line voltage, the waveform distorts horribly. Not unlike a bad current limiter circuit, where when they clamp briefly, the loudspeaker barks horribly, enough to scare ya.

[nevetslab]

Updated schematic, values, DC Voltages & currents

Today I was able to mark up the schematic with component values, xstr types, DC voltages and DC currents thru each stage. I don't yet have designation numbers...the PCB may have those, but for now, I at least have the pertinent circuit information.



Voltages measured with amp powered up at 120VAC. One of the odd findings is the bias xstr is NOT mounted to the heat sink, but left free-standing with all the other small signal xstrs, inches away from the heat sink. The current flowing thru driver stage is frightfully low (4.2mA), as well as the current thru the voltage gain stage/bias xstr also quite low (2.4mA). I haven't monitored these stages as I lower the AC mains yet. I'm just guessing the driver stage needs more current flowing, as well as the voltage gain stage so the circuit can tolerate low line voltage and not have the output crap out the way it is at present. And, I'll remove and move the bias xstr to be on the heat sink as it should be. I haven't yet sent this information to the designer for comment.

2SA940.pdf
2SC2073.pdf
2SA1941.pdf
2SC5198.pdf

[SoulFetish]

Hmmm, my first thought was that you had dropped below the threshold voltage for regulators on the secondary. But, I don't see any regulators.
Any revelations observing the driver stage at low line V?

[nevetslab]

Hmmm, my first thought was that you had dropped below the threshold voltage for regulators on the secondary. But, I don't see any regulators.
Any revelations observing the driver stage at low line V?

In the front end circuits, which I don't have any documentation on, there are a pair of 12V 3-terminal regulators, run off of 24VDC unreg. Those wouldn't drop out of regulation until you're down to around 95VAC or so. No, the issue IS in the power amp circuit. Even if I overdrive the front end, I can see that clipping behavior, while keeping the master down to see the difference.

I just got in to the shop, starting my day, and haven't yet moved to the bench to see what I can find.

[nevetslab]

Replaced Bias Xstr, fashioned hold-down clamp for HS mtg

Starting with the basics, I dug out some MJE720 NPN TO-126 Motorola parts I had on hand to serve as a new heat-sink mtg bias xstr (1.5A/40V part, hfe @ 40), which would fit between the NPN/PNP power xstrs, with a cut-to-fit mica washer. I thought about what to use for a clamp. Spring steel would be nice. Pulled a leaf spring from a broken TRS jack removed from a Roland JC-120, put that in the vise, after cutting off the PCB tip, and bent it over to form an offset-spring clamp that could be slipped under the dual U-Frame clamp holding the PNP outputs down, and with the little dimple in the arm mating with the TO-126 part, that fit in just fine. Greased it up, bent the leads up. Added 26WG Mogami wire, prep'd the ends, added shrink tubing on the device leads. Removed the TO-92 2N5551 bias xstr, wired it up, and powered up the amp again, to see what I needed to adjust.

As I figured, the voltage at the bases of the outputs were now much lower, so I'd have to turn this part off a bit more, trimming the 1.2k B-E resistor in the bias circuit. Didn't take much...7.5k across the 1.2k resistor restored the nominal 450mV potentials at the NPN/PNP bases (re output buss).

It's been running now under load, getting the heat sink hot. Doing what it should......reducing the base voltages at high temp, preventing thermal runaway. I just turned off the signal drive, now the bias is slowly coming back up as the heat sink cools off.



Dropping the AC mains down still causes the output to bark when going into clip, while working fine into clip at 120VAC, so nothing has changed in that regard, as expected. One step at a time.

[nevetslab]

Resuming the chase on this 'simple' amplifier circuit, where dropping the AC mains voltage from 120VAC to 110VAC (8.33% lower) produces rather violent clipping behavior, I first tried decreasing the delta driver emitter resistor (220 ohms), which doesn't pass it's current to the output buss, dropped it down to 47 ohms (60 ohms in parallel with the 220 ohms on board), this made no difference in the abrupt scalloping of the output waveform into 4 ohm spkr @ 110VAC. I then lifted one end of the 220 ohm resistor, and tacked in 47 ohm resistors between the emitters of the drivers to the output buss, and tried again. Same result....works fine at 120VAC, works terrible at 110VAC (and lower).

Pulled up the data sheet on the drivers and outputs. The driver xstrs selected....2SA940 & 2SC2073 are both 4MHz parts, 1.5A/150V parts, hfe of 40 typ. I looked at the specs of MJE15028/MJE15029's. 8A/120V parts, with ft of 30MHz, hfe around 100. I had MJE15032 & MJE15033 parts on hand, basically higher voltage version of the same thing. I removed the 2SA940 & 2SC2073, replacing them with MJE15033 & MJE15032, to see if that made any difference. The bias voltage was now higher, having set it for the other drivers. Checked it at 110VAC, looking for a quick change in results. No such luck....still barking and scalloping the waveform. I saw I had the 80kHz filter on the analyzer, and switched that out, just in case I was missing any oscillation envelope. No difference. I removed the 7500 ohm resistor I had across the bias xstr's B-E junction, which dropped the voltages at the two bases of the outputs to 400mV ea (re output buss). No difference in behavior there either, not that I'd expect any.

I then put a dual gang helipot across the emitter resistors of the voltage gain stage (560 ohms stock value). I dropped those down to 200 ohms, nearly tripling the current thru that stage (and increasing the bias voltage, of course). No difference in the nasty clipping.

There are no base resistors for either the output stage xstrs nor the driver xstrs. I haven't yet tried to remove the base leads from the PCB to try that, as that's a bit tedious.

I looked at the voltage differences between 120VAC mains and 110VAC mains on the output bases, driver stage bases and emitter resistor voltages to see if there was anything significant. All insignificant differences to suggest one stage was on the threshold of starving the next stage.

What am I missing here?

[Jazz P Bass]

Ihe dc voltages look spot on.

Are you sure that it's a power amp issue?

Where does the problem start, looking backwards towards the LTP input?

Could it be a LTP issue?

[nevetslab]

Ihe dc voltages look spot on.

Are you sure that it's a power amp issue?

Where does the problem start, looking backwards towards the LTP input?

Could it be a LTP issue?

LTP? Long Tail Pair? Well, while I haven't looked at the IR drop on the front end collector resistors, the resultant change in their collector voltage shows the expected change in the voltage gain stage current. The amp passes signal, but........with the 8.33% drop in line voltage, SOMETHING is causing serious change in current (I presume) to introduce the radical scalloping of the waveform near/at & above clipping. And, the power xtmr growls in objection. I haven't yet monitored the AC mains current nor current feeding the HV bipolar supplies to see the instantaneous current flow when this occurs. I guess that should be next, since the reading rate of my power analyzer is way too slow to show these instantaneous aberrations.

I've only recorded voltage change differences in the second stage, driver stage (and the resultant base voltages of the outputs as a result) with that minor line voltage change. I'm still waiting to hear back from the designer. I do have some of these amps in our inventory.......haven't yet sent for any to come over for side-by-side comparison. For all I know, they all do it (heaven forbid!).

I'm only monitoring the output waveform into speaker at present. Haven't yet gone looking within the loop at the 1st stage, voltage gain stage, driver stage. The preamp stages are all IC Op Amp based, running off +/- 12VDC regulators, which won't drop out of regulation until you're in the mid-90VAC range, typically.

[nevetslab]

I've connected my current probe on the AC mains line feeding the amp, and also went probing the waveforms at the bases of the driver stage, as well as the bases of the output stage. As everything is inside the feedback loop, I wasn't surprised to find the same waveforms when the scalloping of the output occurred at 110VAC.

But, I did find something interesting. Carefully setting the Sine-random signal at 60Hz, I could drive it well into clip without the nasty aberrations I've been seeing, this still at 110VAC. I then dropped the Sine-random signal to 40Hz, and once again, the severe distortion pattern appears. I forgot to write down my current probe setting, though I think it's 2A/Div. Also at this drive frequency, the amp is pulling more AC mains current than at 60Hz drive frequency.



The first image is 60Hz, showing both the output driven well into clip with AC Mains current via Tektronix A6302 Current Probe/AM503A. The second image is 40hz, showing both the output driven into clip, with the nasty scalloping wave distortion. Notice there appears to be correlation to the mains current and the distortion patterns. Driving the 4 ohm Ampeg BXE-115HL4 15" bass spkr, and with the nasty wave distortion, there's barking and groaning from the power supply, like it's on the verge of blowing up. So far, it hasn't, nor have I blown any mains fuses.

[nickb]

I've connected my current probe on the AC mains line feeding the amp, and also went probing the waveforms at the bases of the driver stage, as well as the bases of the output stage. As everything is inside the feedback loop, I wasn't surprised to find the same waveforms when the scalloping of the output occurred at 110VAC.

But, I did find something interesting. Carefully setting the Sine-random signal at 60Hz, I could drive it well into clip without the nasty aberrations I've been seeing, this still at 110VAC. I then dropped the Sine-random signal to 40Hz, and once again, the severe distortion pattern appears. I forgot to write down my current probe setting, though I think it's 2A/Div. Also at this drive frequency, the amp is pulling more AC mains current than at 60Hz drive frequency.

[ATTACH=CONFIG]54124[/ATTACH] [ATTACH=CONFIG]54125[/ATTACH]

The first image is 60Hz, showing both the output driven well into clip with AC Mains current via Tektronix A6302 Current Probe/AM503A. The second image is 40hz, showing both the output driven into clip, with the nasty scalloping wave distortion. Notice there appears to be correlation to the mains current and the distortion patterns. Driving the 4 ohm Ampeg BXE-115HL4 15" bass spkr, and with the nasty wave distortion, there's barking and groaning from the power supply, like it's on the verge of blowing up. So far, it hasn't, nor have I blown any mains fuses.

Put the scope on the power supply rails on both side of the diodes. I wonder if there's a dry joint / high resistance path there causing excessive droop, especially on the +v rail which seem to be most affected. Low value electrolytics / high esr on the power rails ( don't forget the 100uf ones) would not help either.

A worry is the max output the peak load current is of the order of 10A, more at low frequencies. Taking a combined HFE as 4000 from your figures the drive current needs to be about 2.5mA. Probably more due the HFE rolloff at high currents. The VAS is idling at 2.4mA so there's not a lot of margin there. Maybe try changing those 560 ohms emitter resistors to 330 to see what happens.

[nevetslab]

AC Mains current added to Output Waveforms

Some more output waveform images with the addition of AC Mains current. 40Hz output @ 110VAC clipping has relationship to the AC Mains current, no doubt charging/discharging of the supplies. @ 40Hz Sine output, more mains current is drawn than at 60Hz. @ 120VAC, no waveform aberrations (scalloping) behaviors @ 40Hz, as there ARE at 110VAC 40Hz. Also got some images with 10X scope probe at the base of the NPN Driver stage (fed from upper Voltage Gain stage above bias xstr). As this is looking inside the overall feedback loop, when there is waveform distortion/clipping, you'll see it there too.



The current probe scale is 1A/Div in all images.



These three images are looking at the upper voltage gain stage output feed to the NPN driver base, all at 110VAC, where waveform clipping is occuring. Last image of this series also shows the output waveform. There isn't a direct scaling of that output waveform and scope waveform, as the output is fed from the amp thru an Amber 3501a Audio Analyzer output to the scope. There is a difference in the waveform distortion aberrations at the two circuit locations.

I need to look at this amp driven from kick drum, as that's it's normal application....it being the amplifier for a drum throne transducer. I have a Roland V-Sessions TD-10 Controller here, so will have to fire that up and see if this behavior occurs with the short duration kick pulses.

[nevetslab]

Put the scope on the power supply rails on both side of the diodes. I wonder if there's a dry joint / high resistance path there causing excessive droop, especially on the +v rail which seem to be most affected. Low value electrolytics / high esr on the power rails ( don't forget the 100uf ones) would not help either.

A worry is the max output the peak load current is of the order of 10A, more at low frequencies. Taking a combined HFE as 4000 from your figures the drive current needs to be about 2.5mA. Probably more due the HFE rolloff at high currents. The VAS is idling at 2.4mA so there's not a lot of margin there. Maybe try changing those 560 ohms emitter resistors to 330 to see what happens.

I also expressed concern about the low idle current in the voltage gain stage (2.4mA) as well as the driver stage (4.2mA). I only looked briefly at the output while running the voltage gain stage at 7.2ma. 15mA thru there wouldn't bother me, as well as 15mA thru the driver stage. The drivers are free air, no heat sinks, so I'd no doubt have to add heat sinking.

I was about to look at the power supply rails, though hadn't found any bad solder joints on the board. I'll also look at the supply caps with regards to ESR, using 1khz square wave, 200mA constant current for the 10,000uF 63V caps, and lower current on the 100uF/50V caps at the front end.

Added scope trace of a single power supply rail under full power/clipping....one with V+, one with V- at both 110VAC and 120VAC below:

[nickb]

Ah! I see you did the VAS current in #6 and I missed it. If that were the problem a small increase in idle current would be enough to prove it.

The use of the diodes is a hint that the designer wanted to keep as many volts at lower hum on those early stages as possible. Add check for shorted diode(s) to the list.

[nevetslab]

Ah! I see you did the VAS current in #6 and I missed it. If that were the problem a small increase in idle current would be enough to prove it.

The use of the diodes is a hint that the designer wanted to keep as many volts at lower hum on those early stages as possible. Add check for shorted diode(s) to the list.

Until this last change, I had lifted the 220 ohm emitter resistor between the two drivers, and installed 47 ohm emitter resistors, both tying to the output buss. So, that did raise the driver stage current to 9.7mA. As we've seen, the distortion continues. I had previously lowered the emitter resistors in the voltage gain stage (560 ohms) down to 200 ohms, with the corresponding increase in bias voltage, but, that too didn't eliminate this clipping behavior when running at 110VAC.

I just lowered those resistors to 135 ohms, which dropped the voltage across them to 1.06V, now yielding 7.8mA thru the voltage gain stage/bias xstr, and changed back to a common emitter resistor between the driver emitters, no longer coupling that to the output buss. Used 47 ohms, so the driver stage current, after readjusting the bias to 0.9V across it to 19mA. The results are worse, so made too many changes in one move. I'll go back to the pair of 47 ohm emitter resistors, tied to the output buss, and see if I can still use the higher voltage gain stage current.

Meanwhile, more shop work just came in, which is needed for a rental tomorrow morning. Then to complicate matters, the City of Burbank, who has been ripping up the pavement on the street common to our two properties on both sides of one street, now they've shut down the other street at both ends, preventing ALL truck traffic into/out of our rental depot that services all of our clients (and services me with work!!) Who's running this mess? Trump???!!

I was wondering why the single diode per rail was used to isolate the front end/voltage gain stage from the current gain stages. I do see the waveform distortion is clearly present on the voltage gain stage....of course, it's all inside the feedback loop, so you'd see that everywhere.

I haven't gotten to checking ESR on the supply caps. I'll have to set this project aside again to try and salvage what's left of the month's earnings. More to follow on the project.

[nevetslab]

Using Roland V-Sessions TD-10W Controller Kick Drum signal source

Today, finally back on this odd clipping behavior, I fired up my Roland V-Sessions TD-10W controller, dialed up a decent low-pitched kick drum, and used that with an 8-bar pattern on a drum kit, while only feeding the kick and a bit of snare as the signal source, so this is closer to how this amplifier is used (apart from not having the drum throne with the seat transducer here).

Looked at it both at 120VAC and 110VAC. Same problem. At 110VAC, it comes unglued. The waveform distortion's scalloping of the output waveform appears to be in sync with the AC mains charging current.



I also looked at it with resistive load, and get the same behavior. I've also attached the impedance plot of my test speaker.



It's odd that with 60Hz and higher signal frequency, this doesn't happen, even at lower line.



I checked the ESR of the two 10,000uF/63V supply caps on this amp, and they measure 125mOhms @ 40Hz thru 1kHz, using a 20V P-P square wave, 50 ohm source, measuring the resultant voltage step in the waveform at the test frequency.

So far, increasing the current thru the voltage gain stage to 9mA (from 2.4mA), and 15mA thru the driver stage (up from 4.2mA) hasn't shown any change in this behavior. Mostly there's a trend of higher charging current once below 60Hz signal, and at 110VAC, with the waveform aberrations being in relative sync with the charging frequency. Oddness.