Really? How so?
And assuming for the moment you are right, that does not change my liking them one little bit. They still sound good to me, and that's all that matters to me. They can measure as they like.
The difference between me and many other people, here included, is that I knew from the outset that audio is an illusion, nothing more. It attempts to recreate us being there in person, and it can never do that in absolute terms, but it can do a good job of mimicking it. At any rate, good enough to keep me happy with music I love.
Therefore, I never wasted my time trying for the impossible, searching for perfection, or even tweraking the hell out of my own system. My LP collection and my CD collection are worth more than my hardware. A new piece of hardware will please me, but for delight, it has to be a piece of software.
T, I find it interesting that you would debate with a designer about the designer's design. I think that you overestimate how much feedback is used, because you must subtract the closed loop gain and no special effort is made with this design to maximize open loop gain. For example, the first stage is resistive loaded, and the VAS stage is NOT cascoded. The closed loop gain should be about 25 or so, and this has to be subtracted from the open loop gain to get a feedback estimate. This type of design usually has an open loop bandwidth of 4KHz or so. It is not the IDEAL of 20KHz, but it is better than most in this regard.
Getting a relative amateur to modify an amp like this without a lot of experience is just asking for trouble.
Getting a relative amateur to modify an amp like this without a lot of experience is just asking for trouble.
Specifically, model 6550 (1993), nominally 50/70W 8/4 Ohms, SEPP, has a loop feedback factor of 17 dB.
Model 680 (1999), nominally 85/130W 8/4 Ohms, has a loop feedback factor of 12 dB.
Older models, those near their entry point, such as 620, 630, etc, had a worst case loop feedback factor of 25 dB (model 620). Next up the line, model 630, had a loop NFB of 20 dB, and thereafter, all models were below the 17 dB mark.
Let me remind you all that Denon had a whole series of integrated amps around 15-20 years ago which had no loop NFB at all. And Denon is not exactly a High End company, although they do always have some esoteria in their portfolio.
My own experience tells me you need to be very careful with NFB. Zero NFB has never sounded quite right to me, a little too soft and rounded off for my taste, and how too much sounds you all know from commercial fare. My opinion is to strive for 26...30 dB at most, and that's only assuming your open loop full power response into what you declare as the lowest permissible load (for most, that would be 4 Ohms) is at least 50 kHz or over.
Personally, I feel there's no such thing as an 8 or 4 Ohm load, and I tend to think of 8 as 7 Ohms and 4 as 3 Ohms. Keeps me on the safe side.
John,
I am debating the schematics available in the public domain.
John, I accounted for this.
For the HCA1500, the frontend I make it around 2.6mA per J-Fet (for typical BL Parts), this gives over 20mA/V transconductance per J-Fet or 10mA/V for the differential.
Looking at the VAS Stage we get a gain of around 3.3 to the base of the VAS and the VAS is degenerated with 100 Ohm. The VAS runs at around 7mA so we get around 30mA/V transconductance into the VAS node for each halve circuit.
To estimate the VAS Load, with a FET as driver it is essentially infinity and the VAS Transistors in the HCA1500 I would estimate at around 40K collector impedance (due to early effect) for the PNP Part (that is ballpark, could be higher, possibly a little lower).
But the transistor is degenerated to the tune of around 22dB, so the effective collector impedance will be around 500K for the PNP part and probably around double for the NPN. So the VAS load is around 320 KOhm for the two polarities combined with around 60mA/V total transconductance.
This means DC Gain is at around 86dB at what I'd consider a conservative estimate, with a closed loop gain of 29dB, so we 57dB NFB up to the -3dB Point of the open loop gain.
So we do see nearly 60dB NFB. Cascoding would increase indeed increase it, while using lateral Fet's as VAS may decrease it. Hence a lot will depend on the precise devices in both Driver and VAS positions.
I also got around 4KHz in my "back of envelope calculation".
Note, I am not criticising your design. It is excellent and I learned a fair bit analysing it in quite some detail. I was merely surprised to see such levels of negative feedback in a design from you.
With respect, John, if the amplifier is already stable simply lowering loop gain in the global feedback loop will not cause it become unstable. As all it takes are two resistors (to ground or to negative in). The Mod is easily applied, tested for effects on measured and subjective performance and equally easily removed.
So I would not say it is asking for trouble.
Ciao T
I am debating the schematics available in the public domain.
John, I accounted for this.
For the HCA1500, the frontend I make it around 2.6mA per J-Fet (for typical BL Parts), this gives over 20mA/V transconductance per J-Fet or 10mA/V for the differential.
Looking at the VAS Stage we get a gain of around 3.3 to the base of the VAS and the VAS is degenerated with 100 Ohm. The VAS runs at around 7mA so we get around 30mA/V transconductance into the VAS node for each halve circuit.
To estimate the VAS Load, with a FET as driver it is essentially infinity and the VAS Transistors in the HCA1500 I would estimate at around 40K collector impedance (due to early effect) for the PNP Part (that is ballpark, could be higher, possibly a little lower).
But the transistor is degenerated to the tune of around 22dB, so the effective collector impedance will be around 500K for the PNP part and probably around double for the NPN. So the VAS load is around 320 KOhm for the two polarities combined with around 60mA/V total transconductance.
This means DC Gain is at around 86dB at what I'd consider a conservative estimate, with a closed loop gain of 29dB, so we 57dB NFB up to the -3dB Point of the open loop gain.
So we do see nearly 60dB NFB. Cascoding would increase indeed increase it, while using lateral Fet's as VAS may decrease it. Hence a lot will depend on the precise devices in both Driver and VAS positions.
I also got around 4KHz in my "back of envelope calculation".
Note, I am not criticising your design. It is excellent and I learned a fair bit analysing it in quite some detail. I was merely surprised to see such levels of negative feedback in a design from you.
With respect, John, if the amplifier is already stable simply lowering loop gain in the global feedback loop will not cause it become unstable. As all it takes are two resistors (to ground or to negative in). The Mod is easily applied, tested for effects on measured and subjective performance and equally easily removed.
So I would not say it is asking for trouble.
Ciao T
Hi,
Audacity? Cooledit/Adobe Audition?
Sure. I am still convinced that the dream of lasting peace, of world citizenship, of the rule of international morality may be attained by humankind. Yet when I look around me, everywhere is war. War in the east, War in the west, War up north, War down south - War - war - Rumours of war*, preparation for war, the rattling of nuclear sabres.
So I normally concern myself less with possible dreams of a possible future and concern myself with the same grimy, boring old present. And unless we change the minds of people here in the present, the rest is not worth a rat's rear end.
I agree. It cannot really be a static problem, as under those co nditions any of the Amp's you play with are way past good and evil and should all be "blameless". However, as I found when I build the "Blamless" Amp, it was indeed blameless, you could for example never blame it for providing good sound quality or increasing my musical enjoyment over japanese generica featuring STK "Chip Amp's"...
BJT's have "better linearity" in a sense that parallels "lower THD" and only in some cases.
Looking at the section titled "5. VAS Distortion" in Samuel Groners Commentray on Self's Amplifier Book illustrates that Transistors behave very differently and that at least some FET's have better overall linearity as well as reduced high order distortion.
If it is a Nak CA-5 (not C-5) like this:
The MC Stepup is a single 2SK170 in common gate. I have the service manual if you would like it.
Ciao T
* Adapted from Bob Marley "War"
CA-5Aii to be exact. I would live any service manual. I'll put my e-mail in a PM. I had to have it to sit next to my ST-7. I have never had matched anything.
Audacity? Cooledit/Adobe Audition?
I'll look for these. I have SineGenerator, MultiSine and MDSG. Not very happy with any of them. I actually use the generator built into TrueRta more, but it blows up if you change values too quickly.
It is way too sad what I see around me. Measuring an amp looks like a far more obtainable goal. I always figure, if you don't jump in over your head, you'll never learn to swim.
Thanks again.
Being far more amateur than about anyone in this thread, I am modifying the daylights out of the DH-120 and I picked up a set of rip-off MX50 cards to play with on the bench as they are easier to cut and hack up than the Hafler. My HCA1200 is soundly bolted into my rack being very happy driving my woofers. I intend to keep my paws out of it though I would love to see the schematic. Nothing like seeing the actual implementation in a commercial product to put the designers recommendations into perspective. I did set the bias into spec as John suggested. It was not far off.
In trying to understand the disconnect between reality and simulation, I managed to tweak the various compensation caps in the sim so the feedback remained within 10 degrees 10 to 4meg. This made a very solid oscillator. Finding the center between too small and unstable with too big and unstable, I found if I looked at big magnification at the bottom of a square wave at the feedback signal, the trace was closer to overlaying the input trace and had no little squiggle signs of improper dampening. Now, it see how that translates into the real amp. The phase plot was within 20 degrees to about 2meg, then separated to 60 or so by 4. Spice shows the BW to be quite a bit broader than the amp measures, so I am sure it has reached unity gain long before it exceeds 40 degrees. The models are way too simplistic and I know there are many poles and zeros not accounted for. LT must have much better "internal" models they use for their chip design. Of course, they know what the heck they are doing.
Denon is an interesting company. They always seemed to be in the best value for low end and did frequently build some rather high end stuff, but it never seemed to get respect. No idea if that was deserved. Glad to see they seem to be keeping their hands out of Marantz and Mac, letting them build their own things the way they want/historically did. (DM holdings) Onk was kind of the same. A friend just sent me the manual for a Kenwood KA-9150 he needs me to fix. Surprising how sophisticated the circuit. I would not have guessed matched jfet IPS.
In trying to understand the disconnect between reality and simulation, I managed to tweak the various compensation caps in the sim so the feedback remained within 10 degrees 10 to 4meg. This made a very solid oscillator. Finding the center between too small and unstable with too big and unstable, I found if I looked at big magnification at the bottom of a square wave at the feedback signal, the trace was closer to overlaying the input trace and had no little squiggle signs of improper dampening. Now, it see how that translates into the real amp. The phase plot was within 20 degrees to about 2meg, then separated to 60 or so by 4. Spice shows the BW to be quite a bit broader than the amp measures, so I am sure it has reached unity gain long before it exceeds 40 degrees. The models are way too simplistic and I know there are many poles and zeros not accounted for. LT must have much better "internal" models they use for their chip design. Of course, they know what the heck they are doing.
Denon is an interesting company. They always seemed to be in the best value for low end and did frequently build some rather high end stuff, but it never seemed to get respect. No idea if that was deserved. Glad to see they seem to be keeping their hands out of Marantz and Mac, letting them build their own things the way they want/historically did. (DM holdings) Onk was kind of the same. A friend just sent me the manual for a Kenwood KA-9150 he needs me to fix. Surprising how sophisticated the circuit. I would not have guessed matched jfet IPS.
Sometimes it does no harm to let half informed blabber go on, but when it comes to discrediting some of the best drivers available to the DIY community, the record has to be set straight.
On domes: 1 USD Chinese softdomes are typically ferro fluid dampened, have no resonance chamber, have tiny magnets in course magnet structures. Rather the opposite from Scan Speak and Seas; they are on the fore-front of progress. It is all about construction precision, engineering and material use. The same goes for their larger drivers; very useful. Just have a look at the Zaph site for some independant measurements. Also, generally, their specs are accurate, so you can actually order from the book if you know your own design parameters.
They are pricey though, and you might find cheaper drivers that perform about as well, but you will not know that until you measured them yourself. With Scan Speak or Seas, it is a sure shot. This is not to say that high price and high performance always go together, but in the case of Scan Speak and Seas, they do.
vac
T,
Don't know which "super" Chinese tweeters you are referring to I bought some SB's which are darn good for the mid price range. (designed by former Scan-Speak folks) But, so is the Vifa and Seas in the same range. After playing with as many $50 tweeters as I can afford, I was looking to make the jump. What I think I see in the plots is the higher end drivers are easier to use. Fewer little problems that need extra crossover work. I would GLADLY like to look at other drivers if I knew what to look at. I don't see where the cost is going from the $100 to the $200 units, but I have not had them to play with.
You seem to have a distaste for ferofluid. It is one of those things that can make a driver easier to use. Like everything, it is no magic pill and can by misused. What is your complaint?
dvv,
I was watching "An Idiot Abroad" last night. A series about a very plain gentleman being sent all over the world who does not really want to. One of his comments while being in South Africa, was " they dress different, but they are just the same as everybody else" Why can't we all just understand that! Bad people start wars, good people suffer the consequences. If anyone has a solution, please stand up.
Don't know which "super" Chinese tweeters you are referring to I bought some SB's which are darn good for the mid price range. (designed by former Scan-Speak folks) But, so is the Vifa and Seas in the same range. After playing with as many $50 tweeters as I can afford, I was looking to make the jump. What I think I see in the plots is the higher end drivers are easier to use. Fewer little problems that need extra crossover work. I would GLADLY like to look at other drivers if I knew what to look at. I don't see where the cost is going from the $100 to the $200 units, but I have not had them to play with.
You seem to have a distaste for ferofluid. It is one of those things that can make a driver easier to use. Like everything, it is no magic pill and can by misused. What is your complaint?
dvv,
I was watching "An Idiot Abroad" last night. A series about a very plain gentleman being sent all over the world who does not really want to. One of his comments while being in South Africa, was " they dress different, but they are just the same as everybody else" Why can't we all just understand that! Bad people start wars, good people suffer the consequences. If anyone has a solution, please stand up.
T, I will not argue about your 'calculations' but I think a direct measurement by tvr is in order. All he has to do is to short the 1.8K feedback resistor and note the change in gain. Go to it tvr, and don't break anything! '-)
I would do it myself, by my 1200 blew up a channel and I sent it back to Parasound and we both agreed to toss it out, and I replaced it with an A21. If I had the broken unit, I could have made the test, but now it is impossible.
I would do it myself, by my 1200 blew up a channel and I sent it back to Parasound and we both agreed to toss it out, and I replaced it with an A21. If I had the broken unit, I could have made the test, but now it is impossible.
Couldn't that also be done by applying a signal, measuring the output voltage, then measuring the differential voltage at the input stage?
Yes, concur. Unless he has a fancy low-pF active probe (how I miss the one I had at UCLA) probing is almost certain to disturb things significantly. And if it is a noninverting amp it has to be a differential measurement, even more difficult to do.
WHY not just short the 1.8K feedback resistor, which will give the full open loop gain?. It takes a small clip-lead, only. The only thing that you must be careful with is that the output will change between 30-70dB, and you HAVE to turn down the drive oscillator accordingly, before shorting. The servo should keep the DC offset OK. Just a guess. '-)
Hi,
The record is straight.
Maybe you do not know what is available from the large chinese OEM's? Even more fun is who you tend to meet in their corridors and who has their drivers made there (sorry, signed a NDA so cannot tell).
I must admit that I have a general dislike for typical 1" softdomes. From the cheapest to the most expensive one there is not a lot of audible difference.
The ones I have here have an ND Magnet Stucture and as much chamber as the small magent system allows. Measurements are easily comparable with the best out there. They do not use ferrofluid. And yes, they are 1USD/pcs unselected and bulk packed.
Ciao T
The record is straight.
Maybe you do not know what is available from the large chinese OEM's? Even more fun is who you tend to meet in their corridors and who has their drivers made there (sorry, signed a NDA so cannot tell).
I must admit that I have a general dislike for typical 1" softdomes. From the cheapest to the most expensive one there is not a lot of audible difference.
The ones I have here have an ND Magnet Stucture and as much chamber as the small magent system allows. Measurements are easily comparable with the best out there. They do not use ferrofluid. And yes, they are 1USD/pcs unselected and bulk packed.
Ciao T
Hi,
Nothing "super", just very soldily engineered OEM Parts.
The actual user must add a faceplate (or insert into an opening machined into the front panel) and they actually are also found in some VERY expensive commercial speakers.
I would suggest to look well past the common 1" Softdome genre.
Ciao T
Nothing "super", just very soldily engineered OEM Parts.
The actual user must add a faceplate (or insert into an opening machined into the front panel) and they actually are also found in some VERY expensive commercial speakers.
I would suggest to look well past the common 1" Softdome genre.
Ciao T
- Status
- Not open for further replies.