Hi Wavebourn
You wrote: "apply 1 MHz triangle"
You did mean a 1 MHz square wave ?
So, you just mix that 1 MHz signal with a 20 KHz sine and send it to the amp under test ?
Thanx
Paul
Paul, I apply triangle to see how slopes are equal, and what is their form. No need for additional 20 KHz. If I remember correctly that discussion was related to phase intermodulation. Instead of measuring it directly that is hard and almost impossible I can stretch and extrapolate. Stretch and extrapolate, in most of observations. Find what causes misbehaviour and correct it.
Hi Wavebourn
I've done a search and found in one of your post that you linked that subject to this one:
http://www.diyaudio.com/forums/tube...er-iii-high-end-hybrid-amp-6.html#post1893289
On the scope it's look like a bad square wave, so it's was just a triangle with a bad slope from the output of an amp with phase intermodulation ?
Thanx
Paul
Last edited:
No, that time it was bad square wave. The same trick: expand and extrapolate. This picture looks quite asymmetrically.
This is much better:
Now, let's check on more realistic frequency 50 KHz, 100V P-P:
Last edited:
Hi
Which are best to test phase intermodulation; 50khz triangle or 50khz square wave ?
So we need to look at the best output triangle or square signal for the lowest phase intermodulation, but I've test bad sounding amps showing an almost perfect 100khz square wave on my scope ??
Thanx
Paul
Last edited:
There may be lots of causes of bad sound that can be diagnosed by different symptoms, and phase intermodulation is just one of them.
As I said many times, there are no single recipes to test any amp independently of it's structure.
Hi,
Some of the worst shocks I got from entirely enclosed, fully tested and operational gear came from pure solid state stuff (Computers to be precise). Sure, non where as bad as touching the +B rail in a tube amp, but they where unpleasant enough and by far more frequent...
Ciao T
Some of the worst shocks I got from entirely enclosed, fully tested and operational gear came from pure solid state stuff (Computers to be precise). Sure, non where as bad as touching the +B rail in a tube amp, but they where unpleasant enough and by far more frequent...
Ciao T
Ok,
Hopefully we have covered all battle scars and can get back to listening to ac instead of being buzzed by it ....
@DVV,
In one of my systems when using the Tuner (am) annoying reception noise unless I unplug the power amplifier ( 6 ft away) ...
What gives ...?
@TVR,
Updates on the amplifer , what worked ? Didn't ? .....
Hopefully we have covered all battle scars and can get back to listening to ac instead of being buzzed by it ....
@DVV,
In one of my systems when using the Tuner (am) annoying reception noise unless I unplug the power amplifier ( 6 ft away) ...
What gives ...?
@TVR,
Updates on the amplifer , what worked ? Didn't ? .....
Ok,
Hopefully we have covered all battle scars and can get back to listening to ac instead of being buzzed by it ....
@DVV,
In one of my systems when using the Tuner (am) annoying reception noise unless I unplug the power amplifier ( 6 ft away) ...
What gives ...?
@TVR,
Updates on the amplifer , what worked ? Didn't ? .....
Basically, your tuner may be overly sensitive and possibly not shielded and/or grounded properly, your amp may in fact also be radiating happily away, etc.
I was told by friends who profesionally do RF that grounding any RF device is half the work. It's understandable, imagine the overall gain it needs to have to turn 1 microvolt into just say 100 mV of useful audio signal - that's an amplification factor of (0.1 : 0.000.001) 100,000:1 or 100 dB. Anything on that signal level, let alone more, has a great chance of being picked up. In my Revox tuner, specified at 1V output, the overall gain is another 20 dB up.
Actually, there was a time when I did the scariest test of my life (for me, that is) - I lined up all my tuners and in sequence put them all in the system, and then put my filter straight on top of them. Fortunately for me, nothing came through, with or without antenna attached, from 87.7 to 108 MHz, which is the band we use locally. Wide or narrow selectivity, muting on or off.
While my Revox B760 tuner is not overly sensitive, the old Sony 3950 is an eager beaver, and the Marantz ST6000 is somewhere in the middle.
And my antenna is an active room job, gives a good gain of about 22 dB. These days, I get my signal from the cable TV company - much better. Don't need an antenna any more, in or out.
If you already haven't, try grounding the amp first, see if it goes away, if not, then also ground the tuner.
Last edited:
Happily listening. I stopped half way because it does seem quite improved at least to my classical guitar "metallic" test. It has not had the full wife golden ear test yet, but on quick passing buy, got a "better" comment.
Power supply cleanup of Hexfred bridge, 4X larger main caps with bleeders, replaced the bypass caps, clean up the ground paths etc. Normal stuff. Also removed the bridge and dynaquad garbage, decent binding posts, input jacks. I left the Alps pot I put in for testing but it is basically out of the circuit.
Changed to conventional miller compensation of 100p, removed all other compensation save the 3.3 p in the feedback as the outputs were prone to miss-behave without it.
Moved the gate stoppers to the sockets, 470 and 330 Ohms.
Added 47u bypass on the output sockets
Input cap and feedback DC blocker bypassed with film
Set bias at 110mA. THis is lower than spec but what Exicon recommended and higher that it was. I will jack it up to 145 to see. Set the VAS bias to 3.2 mA
2N5550's are now 5551's and I matched the input and CM pairs to within 1% at 1mA.
I left in the output clamps and funky VAS bias string with protection diodes.
If I were to guess, the change from heavy asymmetric VAS loading, which John lead me by the nose to, probably had a lot to do with it. On paper, it is at least 20 dB lower distortion. Of course, paper lies.
I did not:
TMS (Sim shows great possibilities here)
add degeneration to the IPS CM, (no change is sim, but help with real world?)
change the small transistors. ( only going by specs, no models)
red LED ccs, split IPS and VAS. (Sim shows small advantage)
or
FET IPS ccs, and FBP VAS ccs. ( hunch this is the sweet spot)
The model suggests I would do better at 70p MC, and I have bigger caps and diodes to add inside the rail isolation resistors to help the IPS and VAS rails.
It is stable now, so I have to take a deep breath before I go muck with it again. It was not easy to get the outputs happy.
It has no turn on or off thump, so not pressed to add a relay. I need to study protection a lot more first. I proved the easy way how fragile the circuit is when there is a failure somewhere. Easy, as in easy to blow it up. I must have blown the VAS ccs 4 times. Flakey resistor was the cause. It makes sense to put in modern protection for the speaker and move the rail fuses to the other side of the bridge to protect the transformer. I was thinking about copying Mr. Pass's clever clipping monitor circuit. I have been living with the belief I prefer small amps, but really don't know if I am exceeding their limits and blaming other problems. The old "good 10W"vs 200W argument.
I have all the parts to do this, but I want to test some of it on the MX-50 boards I got. I decided to gut the Aimor chassis I have for the box, transformer and heat sinks. I will build it easier to work on and probe. The Hafler is a pain that way. I need to learn a ton more about protection and recovery before I pull everything out. One trick I thought about was that the sim with the output clamps was very poor. It is back to back zeners. It makes sense that I should be able to replace one in each string with a lower C diode and improve it some. Still, they are technically redundant anyway as the Exicons have them internally. But so did the Hitachi's and E.B. thought it wise to add them on. Hind-sight may be 20-20, but he has a lot more fore-sight than I.
My next step is actually build another set of speakers. If my theory is correct, it is the tweeter at fault, and when I have a good enough speaker, then amps like the HCA 1200 will pass the test with the additional benefit of the detail it has. My theory is the Rotels sound better because they are masking the tweeter problem. With all the bad tweeters, and even worse crossovers out there, that could be a good explanation. They were designed to get the best out of cheaper speakers. High end amps sound poor because they are expecting better speakers without faults to mask. My working theory. Your mileage may vary.
One thing I noticed between the sim of the Rotel and Hafler is that the noise floor and harmonics of the Hafler are lower, but stay about the same across the spectrum. In other words, if 95 dB down 2nd, then 8th is still about the same. Noise floor stays even. But on the Rotel, it falls off at about 6dB per octave. The result is that high order harmonics, which my theory is that they are what is causing the tweeter to be unhappy, are higher in the Hafler.
It is clearly quieter. I perceive it to be more dynamic on my office Kef Q1's, but I also added the Nak CA5 preamp instead of a passive one. I have not put it on the bench to compare the baseline I did as I am busy fixing an old Kenwood 9100 for a friend. I HOPE I found suitable replacement transistors.
Mostly what I am doing is re-reading the books, studying data sheets, and playing with the sims so all the information that has been thrown my way has a chance to sink in. I was comparing the Groner comments to D.S book just today. Re-reading the 4000 posts in this thread for a few details in on the schedule. I instigated some discussions that went quickly above me in the SPICE thread and ANALOG thread I need to glean some more. So much fantastic help. So educational when experienced folks don't agree. Today I have my suspicions, someday I may actually have a based opinion.
I still don't have an explanation for the "delay" from rest differences I measured last year. I need to reproduce that test.
Power supply cleanup of Hexfred bridge, 4X larger main caps with bleeders, replaced the bypass caps, clean up the ground paths etc. Normal stuff. Also removed the bridge and dynaquad garbage, decent binding posts, input jacks. I left the Alps pot I put in for testing but it is basically out of the circuit.
Changed to conventional miller compensation of 100p, removed all other compensation save the 3.3 p in the feedback as the outputs were prone to miss-behave without it.
Moved the gate stoppers to the sockets, 470 and 330 Ohms.
Added 47u bypass on the output sockets
Input cap and feedback DC blocker bypassed with film
Set bias at 110mA. THis is lower than spec but what Exicon recommended and higher that it was. I will jack it up to 145 to see. Set the VAS bias to 3.2 mA
2N5550's are now 5551's and I matched the input and CM pairs to within 1% at 1mA.
I left in the output clamps and funky VAS bias string with protection diodes.
If I were to guess, the change from heavy asymmetric VAS loading, which John lead me by the nose to, probably had a lot to do with it. On paper, it is at least 20 dB lower distortion. Of course, paper lies.
I did not:
TMS (Sim shows great possibilities here)
add degeneration to the IPS CM, (no change is sim, but help with real world?)
change the small transistors. ( only going by specs, no models)
red LED ccs, split IPS and VAS. (Sim shows small advantage)
or
FET IPS ccs, and FBP VAS ccs. ( hunch this is the sweet spot)
The model suggests I would do better at 70p MC, and I have bigger caps and diodes to add inside the rail isolation resistors to help the IPS and VAS rails.
It is stable now, so I have to take a deep breath before I go muck with it again. It was not easy to get the outputs happy.
It has no turn on or off thump, so not pressed to add a relay. I need to study protection a lot more first. I proved the easy way how fragile the circuit is when there is a failure somewhere. Easy, as in easy to blow it up. I must have blown the VAS ccs 4 times. Flakey resistor was the cause. It makes sense to put in modern protection for the speaker and move the rail fuses to the other side of the bridge to protect the transformer. I was thinking about copying Mr. Pass's clever clipping monitor circuit. I have been living with the belief I prefer small amps, but really don't know if I am exceeding their limits and blaming other problems. The old "good 10W"vs 200W argument.
I have all the parts to do this, but I want to test some of it on the MX-50 boards I got. I decided to gut the Aimor chassis I have for the box, transformer and heat sinks. I will build it easier to work on and probe. The Hafler is a pain that way. I need to learn a ton more about protection and recovery before I pull everything out. One trick I thought about was that the sim with the output clamps was very poor. It is back to back zeners. It makes sense that I should be able to replace one in each string with a lower C diode and improve it some. Still, they are technically redundant anyway as the Exicons have them internally. But so did the Hitachi's and E.B. thought it wise to add them on. Hind-sight may be 20-20, but he has a lot more fore-sight than I.
My next step is actually build another set of speakers. If my theory is correct, it is the tweeter at fault, and when I have a good enough speaker, then amps like the HCA 1200 will pass the test with the additional benefit of the detail it has. My theory is the Rotels sound better because they are masking the tweeter problem. With all the bad tweeters, and even worse crossovers out there, that could be a good explanation. They were designed to get the best out of cheaper speakers. High end amps sound poor because they are expecting better speakers without faults to mask. My working theory. Your mileage may vary.
One thing I noticed between the sim of the Rotel and Hafler is that the noise floor and harmonics of the Hafler are lower, but stay about the same across the spectrum. In other words, if 95 dB down 2nd, then 8th is still about the same. Noise floor stays even. But on the Rotel, it falls off at about 6dB per octave. The result is that high order harmonics, which my theory is that they are what is causing the tweeter to be unhappy, are higher in the Hafler.
It is clearly quieter. I perceive it to be more dynamic on my office Kef Q1's, but I also added the Nak CA5 preamp instead of a passive one. I have not put it on the bench to compare the baseline I did as I am busy fixing an old Kenwood 9100 for a friend. I HOPE I found suitable replacement transistors.
Mostly what I am doing is re-reading the books, studying data sheets, and playing with the sims so all the information that has been thrown my way has a chance to sink in. I was comparing the Groner comments to D.S book just today. Re-reading the 4000 posts in this thread for a few details in on the schedule. I instigated some discussions that went quickly above me in the SPICE thread and ANALOG thread I need to glean some more. So much fantastic help. So educational when experienced folks don't agree. Today I have my suspicions, someday I may actually have a based opinion.
I still don't have an explanation for the "delay" from rest differences I measured last year. I need to reproduce that test.
Well done, tvrgeek!
There was a point when I honestly wondered wouldn't it be smarter to simply build a new amp from the ground up, but just as I love my vintage Marantz gear, I guess you love that amp just the same.
Regarding Rotel, they were never my cup of coffee. I always have a feeling of something distinctly artifical going on, at least with their newer crop. Perhaps I am imagining it, I don't know, but they have never done it for me. But that's just me.
As for the Kenwood 9100, if memory serves, it has their custom power packs, doesn't it? If so, you could be in big trouble. The problem is that unlike others (say, technics at al.), their power packs were simply driver and output devices on the same substrate, which was then encapsulated. It's damn hard mimicking that discreetly (usually because of the available space, or rather the lack thereof).
Last edited:
dvv,
You are right on the newer Rotels. They lost it. 800 and 900 series were the best. Again, it could be they are working best for us because of them not doing something so the speakers don't do something even worse. 1000 series I don't like as well.
Right again, the 9100 had their hybrids. Fortunately this one only lost a fet and bjt in the phono board. I can replace the volume pot with an alps and some creativity, but the balance pot is BER. I tried to convince him to just buy a preamp and use the power amp section, but no luck. They bring very good used prices and seem to be highly regarded. He has the matching tuner, which was one of the best ever. It is a case where the owner is in love with it, so I will fix it as best I can. He is actually a competent tech himself but can't work on them anymore because of health.
I would not say I was in love with the Hafler. I like it, invested a lot (about $500 between parts and books) but it has one really big advantage. It fits on a less than 12 inch shelf! What started as trying to reduce the noise for a bench amp has kind of got me hooked.
A news report I saw today mentioned the most counterfeited parts. Analog were the highest category. It confirmed what I thought, seconds being sold as top quality followed by substitutions of very expediencies analog chips with generic junk. I believe the 2N2222a's I got from Digi-Key were seconds sold to them at first quality. Dumpster to your door, thank you China.
I happen to be playing some Harry James right now at "energetic" levels. At least on the Kef's, I don't think it will past muster, but the Q1's are a big step down from the Paradigm 20's. It did a lot better on Julian Bream and Nora Jones. Still running cold, so I guess I can up the bias. Yup. Middle of track 4 "The King James Version" . Not good. Better, but not good. Seas vs. Scan-Speak...... decisions decisions
You are right on the newer Rotels. They lost it. 800 and 900 series were the best. Again, it could be they are working best for us because of them not doing something so the speakers don't do something even worse. 1000 series I don't like as well.
Right again, the 9100 had their hybrids. Fortunately this one only lost a fet and bjt in the phono board. I can replace the volume pot with an alps and some creativity, but the balance pot is BER. I tried to convince him to just buy a preamp and use the power amp section, but no luck. They bring very good used prices and seem to be highly regarded. He has the matching tuner, which was one of the best ever. It is a case where the owner is in love with it, so I will fix it as best I can. He is actually a competent tech himself but can't work on them anymore because of health.
I would not say I was in love with the Hafler. I like it, invested a lot (about $500 between parts and books) but it has one really big advantage. It fits on a less than 12 inch shelf! What started as trying to reduce the noise for a bench amp has kind of got me hooked.
A news report I saw today mentioned the most counterfeited parts. Analog were the highest category. It confirmed what I thought, seconds being sold as top quality followed by substitutions of very expediencies analog chips with generic junk. I believe the 2N2222a's I got from Digi-Key were seconds sold to them at first quality. Dumpster to your door, thank you China.
I happen to be playing some Harry James right now at "energetic" levels. At least on the Kef's, I don't think it will past muster, but the Q1's are a big step down from the Paradigm 20's. It did a lot better on Julian Bream and Nora Jones. Still running cold, so I guess I can up the bias. Yup. Middle of track 4 "The King James Version" . Not good. Better, but not good. Seas vs. Scan-Speak...... decisions decisions
Many seem to have lost their drive of old, almost as if they have given up.
Like most vintage products, meaning the better models, Kenwood/Trio have their own faithful following. Personally, I am not too hot on them, somehow while respecting them, I was never really taken with them. A bit cold for my taste.
As for pots, change for better any and every one you can. That at least is something that never fails when dealing with traditional low cost pots now 30+ years old.
That's how it all starts - first it's recreational, and then it turns into a habit.
That IS a big problem for all of us. I don't even want to start explaining the lengths I have to go to just to make certain I am getting the real deal.
Well then, buy Peerless or Dynaudio. And blame it all on the wife.
China is producing these fakes, but there is no excuse for Mouser to be selling them. They should have /need a better vetting process...
I'm sure they are quite concerned about it. It's a very bad situation. A while back I got some failed systems and after anticipating a variety of likely problems, finally isolated the EEPROMs as the culprits. I contacted the negotiant with the China factory, suspecting counterfeit parts, and found out to my surprise that the US liason had bought them here and shipped them over there!
- Status
- Not open for further replies.