[IpsilonSound]

I tried replacing my opamps with LM6171 and THS4061(81) which are more superior(IMO) to the AD8610 chips installed in my PPA. The newly installed opamps(LM7161 or THS4061) work great in PPA having HAL5002 buffers but don't work with Larry's descrete diamond buffer for some reasons

The sound is very distorted, the opamp oscillate and when turning of the power, the amp gives a loud pop noise.... I haven't tried Tangent advice on how to work with cranky opamps, but it's kind of wierd, since the power, gain and bypassing for the opamp is pretty same on both setups, it's the buffer which makes a difference. Plus most of Tangent's recommendations have already been implemented on PPA.

I would like to install the THS4061 amps on my PPA, the difference isn't subtle. The chips sounds very transparent and crystal-clear, the AD8610 sounds a bit dead and veiled compared to it.

Did I mention that my Grado-80 phones were killed by this? Obviously too large offset giving loud thump when switching power... I should be more careful next time

Anybody knows why it doesn't work with the descrete buffer?

 

[Glassman]

it's a strange behavior, but luckily, we haven't run into this kind of problem with my discrete buffers.. Triad had originaly problems with OPA637, but we haven't, maybe our buffer will work even with those opamps, you'll have a chance to try it soon.. you should try increasing the resistance between opamp output and buffer input, replace the 1K resistor with 10K and see what happens..

 

[xtreme4099]

my triad seems to work fine with 637s i think it depends on the setup ...

 

[IpsilonSound]

I think that the most probable cause of this is the difference in capacitative loads. The DDB from Larry adds more capacitance on the amp load. The AD8610 is stable with capacitative load over 1000pF and THS4061 needs adjustments, like increasing the resistance between the opamp output and input of the buffer(suggested by the datasheet):

 

[morsel]

Quote:

I haven't tried Tangent advice on how to work with cranky opamps...

Strictly speaking, this is not advice, but a set of mandatory requrements. It is not about luck or strange behavior. The notion of using the LM6171 without a feedback capacitor or the proper resistor values is absurd.

 

[IpsilonSound]

Quote:

Originally Posted by xtreme4099 my triad seems to work fine with 637s i think it depends on the setup ...

I also had no problem with OPA637's on mains and DDB from Larry provided that the gain was > +5.
The THS4061 and LM6171 chips came from another league, they are 180mHz ultra-fast, very low distortion amps which are even suitable for video applications, let alone audio, where they outperform their audio-dedicated amps. Such fast opamps are not recommded for socketing, since the socket legs may create additional capacitance. Provided the setup is configured properly, the PPA and diamond buffer should be working great and on the new better level. Anybody can try the THS4061 without DDB and notice the improvement in the sound, it's not small. The THS4061 can be ordered as samples from TexasInstruments. But my advice - try it first with a cheap set of phones, since the offset may be too large to destroy your headphones, virtually 'melting' the membrane like in my case(it was a combination with DDB though). Since the PPA doesn't limit bandwidth, except for the ground channel, only God knows what frequency may get thru these chips.

 

[ppl]

I have not had the LM6171 work with the HA-5002 since the DC offset changes with volume control posistion. Also only a few PPA dont work with the OPA637. Mine has OPA-637/627 for a few mounths now without issue except one time when a power supply cap became defective. and as extreeme says his did also however he likes the LT1028 Better. Regardless of what your PPA configuation is i do not recomend any non fet op amp in the PPA except the LT1028 all others have showen changing DC offset with volume controll posistion. Now Glass man claimes to not have had this issue however i belive he has not tried these in as manny configuations. i spent mounths in R&D on my DB so i know a thing or two about what will and what will not work. But Hay if the 5002s work for ya use them but i would sugest you measure the DC offset vs different volume controll posistions and see for your self YMMV

IpsilonSound How do you figure the DB adds more capacitence than 12 paralell HA-5002's since the input transistors are low noies types and low capacitence i doubt that this is more than 12 IC buffers, while the HA-5002 is devoid of input capacitence data this makes me belive it might be High as when a Manufactur omitts a spec its most often because it is not as good as the competition like the BUF-634 output impedance spec.

 

[IpsilonSound]

PPL, I only assumed that the DB was more capacitative than HAL5002's, since the voltage-feedback type of opamps like THS4061 or THS4081 oscillated with DB, where they worked great with the chip buffers. The first cause of oscillating THS chip according to the datasheet is capacitative load.
Not that there's something wrong with the DB, it's a great update to the PPA, but there are some compatibility things with a few chips. The same thing happenned to me when I tried OPA637's with the descrete buffer, only increasing the gain from 4.7 to about 7-8 eliminated the problem completely. Yes, it's the OPA637 chip which is stable only with a gain of 5 and more.

I managed to fix the oscillation problem with DB and THS4081 by raising the gain (R4=6.64k) and also putting a 100pF mica cap between the opamp (-) input and output. The cap reduced the bandwith of the signal in the ultra high frequency region and the oscillations are completely gone. Also ultrafast chips like THS needs to be soldered directly to the pcb to reduce the capacitance, which I also did. Anyone knows where to get the formula calculating the parallel capacitor value affecting the frequency of the signal? I've seen it somewhere before. Curious to know what the cutoff of the bandwith with 100pF cap.

Measuring the offset with changing the volume will be the next thing tomorrow, it's too late already. I only measured the current from the chip with no signal and it was around 80mA, which was correct according to the manufacturer's specs. The AD8610 outputs 45mA, for example. The THS4061 shouldn't be used here, since it gives 160mA of current and takes twice more power, gets more warmer.

Anyway, I'll do more critical test with another set of phones in the near future. Meanwhile, the difference between AD8610 and the voltage-feedback THS4061 chip is very tangible, more clear, pristine sounding music.

If the PPA design is reviewed with the possibility of using voltage-feedback ultrafast opamps, it will be a great update, I think. The sound difference is big, as well as the reduced level of distortions.

It may sound funny

, but I realized why my Grado phones died yesterday, it had nothing to do with the nature of THSxx chips. I tried to test it when switching opamps and forgot to put the opamp for the ground channel, only one channel had THSxx chip installed, when one headphone driver started sounding quieter than the other. I did the same thing twice again, forgot to put the opamp for ground channel by mistake, and noticed that one of the phone drivers got real warm playing really distorted sound. But it's ok, I turned the power off real quick

Good lesson - never test PPA without proper opamp installation!

 

[morsel]

That is capacitive load, capacitative load is how many McDonald's french fries you can stuff into your mouth at one time.

 

[IpsilonSound]

Quote:

Originally Posted by morsel That is capacitive load, capacitative load is how many McDonald's french fries you can stuff into your mouth at one time.

Good analogy, Morsel

, though, I don't like mcdonald at all, I prefer healthy food, you know.

 

[tangent]

Quote:

Anyone knows where to get the formula calculating the parallel capacitor value affecting the frequency of the signal?

It makes a simple RC low-pass filter with the R in parallel, so the corner frequency of the filter is f=1/(2*pi*R*C). If you put 100pF in parallel with a 10K R4, for instance, the filter would be down 3dB at about 1.6MHz, and fall 6dB per frequency doubling beyond that point.

Quote:

never test PPA without proper opamp installation!

I prefer "never test a PPA with your good headphones".

 

[morsel]

Quote:

never test a PPA with your good headphones

Thank goodness he used the Grados, then.

 

[ppl]

Well lets see how many grammar errors Morsel can find in this reply, First off the Diamond buffer and the PPA was not and still is not designed for these fast op amps. There is no ground plane and this is mandatory for these chips, also the resistor values around the op amp and the input are several times larger in value than recommended in the data sheet. In addition a different layout around the op amp proper would be required as well. The HA-5002 is a 110 MHz device and is not intended to be within the feedback loop of a 200 plus MHz op amp without phase lead compensation. This is something I wish to avoid so I recommend op amp that are slower than the Buffer to avoid instability. These issues in totality add up to a no win situation.

These issues are classic feedback theory and just because a few examples worked that are in contrast to accepted feedback theory, is in no way going to make me recommend these unscientific misapplication of a given op amp as clearly spelled out in the data sheet.

With regards to the DB not working with these op amps, I am not surprised that it don’t and the reason is not the capacitance see by the op amp as the 1k resistor between the op amp and the buffer completely isolates the op amp from any buffer input capacitance. So what would make the IC buffers work with a given device and not with the DB is again layout. Anyone that deals with RF design knows that the area of the loop affects high frequency stability. In the Triad case this is a module that plugs in to where the IC’s once were, this is now a lot larger area for high frequencies and allows a greater risk of input to output coupling from the input of the PPA to the output. Another area if concern is the input of any Emitter flower become inductive and on concert with the transistors capacitance form an RC resonant circuit that will be excited by any radio frequency stimulation. The solution for this is bas stopper resistors on the output and Driver devices the output already includes this as part of the Class AB bias network however the buffer input transistors do not.. I did not place these resistors on board due to lack of room and the fact the the 1k resistor in place on the ppa board proper would perform this function. I do think that maybe the distance from the base connection of the input transistors to this 1K resistor could be too long to allow proper cancellation. The input transistors have a Fy of over 400 MHz at the operating current I use so foil length is critical. The Module is mounted on stand off and this add the problem.

 

[jerry1130]

Quote:

Originally Posted by xtreme4099 my triad seems to work fine with 637s i think it depends on the setup ...

Mine too, except when driving some low impedence headphones like ER-4p or R10 lol, I can hear cracking sound coming from the amp, the problem goes away when I convert the 4p to 4s or plug in a higher impedence phones. One thing I don't get is that CD3ks work perfectly.

 

[IpsilonSound]

Quote:

Originally Posted by ppl First off the Diamond buffer and the PPA was not and still is not designed for these fast op amps.

Well... unfortunately not, but if(in the future) changes are done to allow for fast opamp implementation, the PPA will be a totally new level of the headphone amplifier. A simple search on the net for 'voltage-feedback amplifiers for audio' returns proof why so many engineers already look at them as a highest-quality audio solutions possible. The sound difference is REAL and BIG compared to jfet audio opamps and you know it too, Phil

.

After all, why the CMoy amp was initially designed for using fast opamps like LM6171 and the higher-end PPA is restricted to slower opamps, shouldn't be the other way around?

Quote:

Originally Posted by ppl There is no ground plane and this is mandatory for these chips

Right, it's part of the recommended design consideration for any fast opamp running full-speed, and would be ideal for PPA+fast opamp combo, of course. But in our case we don't and shouldn't use the opamp's full speed(175mHz for THS4081), the mica 100pF cap is placed right on the chip reducing it's bandwith to 1.6 as Tangent said earlier(f=1/(2*pi*R*C). Therefore the bandwidth reduction plays the key role in preventing the chip from oscillating at lower gains.

Quote:

Originally Posted by ppl the resistor values around the op amp and the input are several times larger in value than recommended in the data sheet

The values of 1.3k resistors in the datasheet are just given as an example to demonstrate how to balance the offsets on its two inputs of the opamp and avoid the high output DC offset, as again, proved by Tangent in the article called "Working with cranky op-amps". The datasheet doesn't prevent us from using the resistor values which are more applicable to our case with PPA(varying input, buffers, biasing to ClassA, etc).
Using the Tangent formula r2=((r3*r4)/(r3+r4)) and the standard values taken from PPA AND the datasheet the equation works equally: 4.3=((4.32*1000)/(4.32+1000)) or 0.6=((1.3*1.3)/(1.3+1.3).

Quote:

Originally Posted by ppl These issues are classic feedback theory and just because a few examples worked that are in contrast to accepted feedback theory, is in no way going to make me recommend these unscientific misapplication of a given op amp as clearly spelled out in the data sheet.

A huge deal of brilliant scientific discoveries were brought into the world by unscientifically misapplied and often crazy experiments, those very ideas that were initially disapproved and critisized as deviation from the commonly accepted theories and dogmas.

Quote:

Originally Posted by ppl So what would make the IC buffers work with a given device and not with the DB is again layout.

The bandwidth reduction cap is a good workaround until the layout is different. It now works great with both DB and IC, not scientifically approved though

.

Quote:

Originally Posted by ppl The input transistors have a Fy of over 400 MHz

The THS4081 is 175mHz which is below the buffer's speed. There are even faster opamp chips, but with more problems using them, like bigger DC change, more power consumption, that one is fast enough already and can be successfully implemented.

The DC changes with volume and there's probably no cure for that. I measured around 260mV at full volume with r4=6.64k, and around 15mV at 9 o'clock. No one will likely be listening to PPA at full volume for a long time, but how may such relatively higher voltage affect the operability of DB in the long run? The channel output is the same, of course, since it's regulated by the DB and unaffected by the opamp change.