Pesky thin traces. I experienced a similar problem with a product I was designing for a client. I had ringing at around 150Mhz. I doubled the thickness of the PCB tacks. Problem solved. Because I needed to source several hundred ma, I couldn't afford series resistance.
So are we saying that if it actually makes an improvement, then it works by damping an HF resonance in a poor PCB design? A low value resistor would be cheaper, and more linear.
The sad thing is that some people might look at that circuit and think it is an emitter follower!
The sad thing is that some people might look at that circuit and think it is an emitter follower!
Yes, I like to do something similar - all my op-amps have a 22 Ohm resistor in series with the supply pins as close as possible, and then a 100uF from the op-amp supply pins to a 'decoupling' ground. It may well be with the mod made by Chris im provesthe sound because its dampening ringing on the supply rails. But, I would argue that a gyrator, or well executed rail filter of the type described above would do a very good job. Few people realize that as little as 10cm of copper track starts to make itself felt at 100KHz. For a wideband op-amp without suitable coupling, this is a sure fire way to get sub-optimal results.
Well, no-one is questioning if it works, just your explanation of why.
You seem to be making a typical conclusion that if you understand it to be an emitter follower (which it certainly is not) because the definition of it shows something favorable to solve your problem, then if the problem is indeed solved, you conclude it indeed IS an emitter follower.
As DF96 says, more reading is required. An emitter follower referenced in your article follows the very AC signal you removed.
In general, a lot of problems with op-amps come from people not understanding the word 'decoupling'. A capacitor in parallel with the power supply pins of the OPamp does not work alone, in fact it uses the parasitic properties of the power supply wires or traces to form a filter. I.e. it AC decouples the currents flowing in and out of the power supply pins from the rest of the power supply 'grid', keeping the current loops close to the chip, by providing a lower impedance path for them the higher the frequency. However, most of the parasitic properties of wires and tracks consist of inductance. This forms a resonant circuit with the decoupling capacitor, so resonant ringing is possible. With old opamps this was not such a problem since they were incapable of reacting or indeed amplifying at these frequencies. Also, becasue the actual resistances and inductances are fairly low in value, the Q factor of the resonant circuit is high, and the corner frequency also.
Adding a resistive element in series with the power supply can actually improve the situation in some cases, but adding a discrete resistor (usually in the 10 ohm or so range) and using proper decoupling caps, can make all the difference. Not only is any resonant circuit damped, but now you can use fairly large caps in order to keep current loops around the opamp short even at quite low frequencies - in fact, the dominant behaviour will be that of a RC rather than LC filter, and a RC filter is incapable of ringing, so you can safely dimension it to have some efficiency at fairly low frequencies.
It should be noted, however, that there are different types of decoupling, depending on where the actual power and output currents go. If there is a 'common reference point' such as ground, the load that the opamp sees and the decoupling should be returned to this point, and in fact you get a 'local' start ground of sorts. The point is to keep AC currents out of the ground and power lines wherever possible, or at least keep any common path of these currents with other currents as short as possible - one of the techniques to do so is deliberately inserting a resitance into a power (and less often, ground) line, in order to be able to introduce a bypass/decoupling path of lower impedance at frequencies of interest, hence aformentioned 10 ohm resistors and the like.
Last edited:
High ESR capacitors will usually fail in a rather spectacular nature when used in a switchmode PSU 🙂
To me the setup looks like an opamp sandwiched between two cascodes, the only thing being different is them being referenced against the supply rails with a built-in dampening through the base resistor together with the input capacitance. At high frequency current swings the opamp inner sees a more constant voltage if the supply line is prone to sagging or otherwise not properly decoupled. This topology will probably do very good in addition with the standard opamp decoupling of 100nF MKTs with a 10uF electrolytic for each device per rail.
No, this is just a non-linear resistor in series with the supply. Whatever droop the supply already has will be maintained, plus what ever extra drop is introduced by the resistance.MagicBox said:
If he put a cap to ground from the BJT base then things would be very different. He would then have an emitter follower, at least for AC.
Hi
All the information to try it yourself is provided at my blog site opamp - opamp construction, so my subjective can be yours as well. I look forward to you trying it, and with the courtesy having done so, of a reply to the forum.
All the information to try it yourself is provided at my blog site opamp - opamp construction, so my subjective can be yours as well. I look forward to you trying it, and with the courtesy having done so, of a reply to the forum.
The 2 transistors will be totally saturated with the resistor values shown and say 6mA quiescent current for the AD825, so definitely won't act as any sort of emitter follower, just a bit of non-linear always turned on diode-like behaviour. If there's any black magic involved as far as sound quality improvements, it's the removal of ringing in any supply filtering as discussed by others, and I wonder is there some part cancellation of distortion generated in the opamp? i.e. 2 distortions cancelling themselves out. Who knows on that one.
Im sure if you tried this idea out you yourself would know. Theory is very good as long as you apply the right theory. Im guessing you have applied the wrong theory as you still cannot see the benfit. I have listened to the effect this idea has and i can say the improvement is very good.
Forget black magic forget wrong theory and have a listen . Oh and before you ask i have not applied theory to this idea and i dont have a theory on it .. Listen Man it works .
Regards Ian
A good theory is only proven by practise less thinking and more doing thought alone has no use whatsoever in this practical world
I agree with this description of a turned on diode provided by sbrads, firstly thank you to all who provided responses.
The transistors produce best performance between 5ma and 7ma, I have tried also at 10ma but op amp case work with transistors underneath get appreciably warm... and sounds great. So at 6ma they act as a always turned on biased diode which is a sensible bias point. I do not think the same performance can be achieved with resistors or plain diodes, rather having a fixed bias is the way to go and Bc856 and Bc846 appear to be good transistors for this task.
If there is cancellation of distortion generated by the opamp, the use of transistors connected this way provides very consistent result. Every op amp I have modified this way returns the same result. Better stereo which brings so many nice benefits and bass also improves.
Exactly how they interact with the op amp remains fascinating, and well worth while for you to try
if getting the most from your op amp is required.
Cheers / Chris
The transistors produce best performance between 5ma and 7ma, I have tried also at 10ma but op amp case work with transistors underneath get appreciably warm... and sounds great. So at 6ma they act as a always turned on biased diode which is a sensible bias point. I do not think the same performance can be achieved with resistors or plain diodes, rather having a fixed bias is the way to go and Bc856 and Bc846 appear to be good transistors for this task.
If there is cancellation of distortion generated by the opamp, the use of transistors connected this way provides very consistent result. Every op amp I have modified this way returns the same result. Better stereo which brings so many nice benefits and bass also improves.
Exactly how they interact with the op amp remains fascinating, and well worth while for you to try
if getting the most from your op amp is required.
Cheers / Chris
Last edited:
And in this case the right theory is ?madtecchy said:
It seems to me that, depending on the current gain of the BJT, we either have a current-biased transistor or a transistor in collector saturation. The latter would give a lower impedance and so maybe less extra distortion.
If this really changes the bass then it must be a thermal effect in the BJT. How this finds its way past the opamp PSRR is for others to discuss.Chris Daly said:
Yes I have tried ,and neither resistors or plain diodes achieve this audio improvement.
Cheers / chris
Cheers / chris
See if you'd get even a better performance if you were to actually use a cascode circuit. The bases not wired directly to the power rail, but say, a fixed reference under the power rail, say 5%, allowing for a bit of Vce to exist across the transistors.
The way of biassing, saturating the transistors is what makes me think the setup acts as a half arsed cascode eventhough it's not a real cascode circuit.
The way of biassing, saturating the transistors is what makes me think the setup acts as a half arsed cascode eventhough it's not a real cascode circuit.
I guess you are asking me what the right theory is. if you care to stop looking for an argument in posts you will have read i do not have a theory!End of. As i like a good argument1 you are not my chosen to argue with in any constructive way. In any way for that matter . also the post was not in reply to you in any way. Now if you would like to be littel me in any way i can tell you now you will be wasting any time and effort you expend . Now back to the intresting thread hey. Oh by the way im sorry for putting a smudge on this great thread . Df had the option to keep this private so he is half to blame too.
Last edited:
MagicBox,
In my book, that would be an emitter follower and very much in line with adding a capacitor to the base as DF96 has mentioned.
Hope this helps
-Antonio
- Status
- Not open for further replies.