Sound Quality Vs. Measurements

Status
Not open for further replies.
Cleaned up DH-120 with the bridge and dyna-qyad removed
 

Attachments

  • DH-120-2.jpg
    DH-120-2.jpg
    126.9 KB · Views: 218
Hi,

I was not going to ask about that diode until I understood a lot more.

Looks to me as if it limits the voltage across the current mirror.

I would probably give the mirror some degeneration AND I would look at adding J-Fet buffers to the existing LTP. Use a 2SK389/LSK389 with a 56K resistive tail. In any case, I'd probably up the emitter degeneration resistors quite a bit. Especially with a CM load the degeneration only improves linearity, gain is only minimally reduced.'

Alternatively going for J-Fet inputs, at a minimum with zenner diodes in the drain leads to keep within the voltage limits, preferably fully cascoded may be better.

I'd also look at the VAS loading RC (C3/R11), it seems a bit brutal. It may be smarter to use a higher value and return it to the inverting input (See Bob Cordell).

C9 & C14 can probably use to fit the largest value you can physically fit (a few 1,000uF will do no harm and a Diode in series with R17/R22 may then be beneficial.

Q6, Q7 & Q9 are quite critical to sound quality, I think you can better than what is in place. Samuel Groner's commentary on Self's Power Amp Design Book can serve as beacon (another piece that deserves JAES publication and self publication). You could even use a J-Fet or Mosfet for Q6.

Q10 & Q11 are up next, Goldmund uses some very low Cob transistors there, for good reasons. These used to be "Video Transistors", with very high voltage ratings and very low Capacitances and high Ft, it seems most manufacturers obsoleted these parts with the demise of CRT TV's...

C1 and C4 would be next on my list, C1 for shorting out if a J-Fet (buffered) input is used, and with Fet buffered input you could use a "trick" circuit to deal with C4 - use a 1M resistor from the output to -Input and then a 1uF Film Cap of high quality to the point marked "B" on schematic, with the original connection disconnected.

Your DC Blocker capacitor is now inside your feedback loop and thus it's sonic impact is reduced and you can use a film cap. If you like a little "soft/warm" flavour for the Amp find Polycarbonate Cap's for this, I loved the old Siemens stacked film polycarbs for modding SS Gear, shame they are history.

If you use MKP you will have a bit more clarity, MKS will be even warmer but perchance a bit too soft. Inside the feedback loop the differences are quite a bit reduced.

My final point of attack would be to look at all these clamping diodes (D5, D6, D10, D13, D14, D15 - the others are voltage references) and look for the lowest reverse biased capacitance diodes you can fit.

Have a look at all these compensation cap's. If you can find them, my recommendation is non-magnetic mica of NP0/C0G Ceramic.

As a last item, the rail fuses may warrant attention, both fuse holders and fuses. Not that I'm suggesting expensive audiophile fuses there, but not all fuse holders and fuses are created equal. I would look for magnetic materials, good contact pressure etc. first.

Ciao T
 
Notes taken. (5 pages from this thread) Thanks. I am avoiding big circuit changes until I understand it a lot better. Baby steps. When I get the Fairchild parts, I'll probably go ahead and put 100 Ohms in the CM, just because I'll have the transistors out. For now, study.

I also need to take time out and get the process for measuring IM down. I think I can do it with my existing PC tools.
 
diyAudio Member RIP
Joined 2005
Thorsten makes many good points. I lament the passing of the CRT driver parts mentioned as well. John Addis told me about a current source he'd designed using a Philips (pre-NXP) 100V low-output-C part in a Larson* super-alpha pair with a 5GHz auxilliary transistor, with a 100V breakdown and a broadband output capacitance well below a pF, and sadly the main part is long gone. And Groner's articles are quite worthy of study.

But I would proceed with caution --- as you well know you're playing with fire here. I can't for the life of me see how that one particular diode helps, but EB surely put it in for a reason. And it seems the unballasted (no emitter resistors) current mirror's only virtues are pulling closer to the rails and being fast, but entail a noise disadvantage (see Groner for discussion).

Brad


*the circuit that feeds the main transistor base current back into its emitter via a complementary common-base stage, which I recently discovered dates back to at least 1957 as applied to common-base amplifiers for telephony by one Dr. Frank S. Boxall, work that apparently thereafter went mostly unnoticed. Larson at Tektronix filed a patent app in 1965, Baxandall and Swallow wrote about its application as a current source in 1966, and it's most-known today as the Baxandall super-pair (I'm sure none of them were aware of Boxall). Boxall called it the Feedback Compound Transistor (see for example Semiconductor Products magazine, Sep./Oct. 1958, especially Fig. 9(c) ).
 
Hi,

I can't for the life of me see how that one particular diode helps

It is a single ended VAS. At clipping it can be driven to quite a decent current, more perchance than it likes.

The diode and other halve of the CM limit the base voltage of the VAS Transistor to two diode drops, say 1V, this limits the current. Nowadays often a protection transistor is employed instead...

I think I like Erno's diode more better. There are also baker clamps and other clamps, they may perhaps be not necessary and may omit them, but Erno was clearly going for "Belts, Braces, Flack Jacket, Ceramic Lined Helmet and a sandbag wall" in his fight to make sure the Amp stayed reliable (nothing for SQ) and recovered quickly and cleanly from any clipping (which does have an impact on SQ though only if we are clipping the amp).

Ciao T
 
Your DC Blocker capacitor is now inside your feedback loop and thus it's sonic impact is reduced and you can use a film cap. If you like a little "soft/warm" flavour for the Amp find Polycarbonate Cap's for this...

So the difference between capacitors which is so small it is dismissed by many, and as far as we know unmeasurable, is also reduced inside an amplifier's feedback loop, and yet still gives a "warm" flavour..?
 
Hi,

So the difference between capacitors which is so small it is dismissed by many,

Many dismissed the ideas of the Wright Brothers, it did not make them "(w)right" now, did it.

and as far as we know unmeasurable,

Please do not use "we". You may not know how to measure differences between capacitors, I do.

Heck, D. Self does.

He even knows how to measure capacitor break-in (another thing dismissed by many).

is also reduced inside an amplifier's feedback loop,

If negative feedback works correctly and the capacitor is placed inside the feedback loop, yes. In addition we might observe that AC current through the capacitor is minimised in the suggested scheme.

If the DC blocking cap is placed the traditional way (as part of the gain setting circuit of the feedback loop) not only is it exposed to quite high signal currents, any of it's non-linear effects are not addressed by the feedback loop.

and yet still gives a "warm" flavour..?

Subjective personal experience.

Take or leave it (all of it), I could not care less. I am trying to give someone who wants to modify an amplifier some pointers. I have better things to do than to get into a futile debate about capacitors with people who do not even know how to measure differences between capacitors.

Ciao T
 
diyAudio Member RIP
Joined 2005
Hi,



Why would it distort if it is reverse biased (other than capacitance)?

Ciao T

It looked particularly yucky in open loop with very light loading, even with a very small voltage across the diode (at quiescence, according to my model, it has a very small forward bias, so any signal swing is modulating the capacitance a lot --- but then the voltage doesn't have to swing much to drive the following stage). But I did some more accurate sims with the following stage appended, while still not closing the loop around the whole thing, and it doesn't look nearly as bad with the loading on that node.

If I had some better DMOS models I might go ahead and simulate the entire amp just for fun.

And putting in emitter resistors would vitiate the clamping function, so it's understandable why things wound up this way, one of many design tradeoffs.

Brad
 
I have better things to do than to get into a futile debate about capacitors with people who do not even know how to measure differences between capacitors.

I'm sure you can measure a difference between capacitors - it would be surprising if there were no difference, and if there was a difference I think I could do that without a problem, thanks - but in this application and position in the circuit, what I should have said is, how measurable is the difference in 'warmth' of the sound? It sounds as though it should shout out from the frequency response or other measurements. Sorry for my ignorance. How big is this measurable difference?

(Not expecting a reply of course!)
 
Cleaned up DH-120 with the bridge and dyna-qyad removed

Nice amp, well designed. Each janitor uses own brush, and brushes differently, so opinions may be different, but if to follow all of them you will come to many different designs, not necessary better than this one. This design solves one problem that the author obviously concentrated on: clipping behavior. And it is definitely well breadboarded, I doubt that computer simulation can be used to improve it.
 
Hi,

It looked particularly yucky in open loop with very light loading, even with a very small voltage across the diode (at quiescence, according to my model, it has a very small forward bias, so any signal swing is modulating the capacitance a lot --- but then the voltage doesn't have to swing much to drive the following stage).

Yup. I find whenever I see something in a circuit that has both a decent reputation and is designed by a competent designer, it pays to double & triple-check.

If I calculate right, the diode has 62mV forward bias at quiescent. If I calculate right the Amp needs around +/-100mV on the base of Q6 (including degeneration) to swing rail to rail and the input impedance of the VAS is around 3KOhm.

I'm coming with approximatly megaohms for a low leakage rectifier (no schottky), next to the 10K or so of the undegenerated current mirror.

And putting in emitter resistors would vitiate the clamping function, so it's understandable why things wound up this way, one of many design tradeoffs.

The CM degeneration would not change the clamping point materially enough IMHO to endanger the Amp... It would worst case add around 88mV to the clamping voltage for 100 ohm CM degeneration.

Ciao T
 
Administrator
Joined 2004
Paid Member
Hi Brad,
That "epoxy" is known by some as "Sony bond". It's purpose was to keep parts from moving in the solder bath step. Sadly, this adhesive is corrosive and conductive to varying degrees. Heat will discolour it and increase conductivity, and it is hygroscopic to some degree, also increasing conductivity. Nice. It's use was continued long after these negative characteristics were generally known. No service advisories were created by anyone to the best of my knowledge either.

Removing this adhesive is the #1 job of anyone working on equipment where it has been used. Light tan in it's "nice" state, turning to a dark brown to almost black in it's most disruptive states. This stuff has made D/A converters distort heavily, wired remote control inputs non-functional and halted uPs inside equipment. It has eroded both copper traces and component leads - though not as badly as Coke or feline urine (the most destructive stuff ever).

In my personal opinion, early NAD products are not a bargain! They are not designed very well and the execution of the design is very poor. I've serviced many in my career and found they have cornered the market on odd faults. The design of the amplifier has a thermally sensitive drift due to the singleton input stage and semi compensated current injection for offset. The output "protection" consists of a self resetting breaker that suffers from variable contact resistance. All in all, you would be well to avoid this stuff, especially since so many good examples of engineering are available at a cost about the same as a used NAD.

There is one thing that the NAD team has done exceptionally well. The advertising campaign.

-Chris
 
Status
Not open for further replies.