Body
Ole,
An amp built with super high Q inductor current sources, sounds like RF
I don't think my tweeters would enjoy that either or the transformers in
the electrostatics.
Her amp is interesting yes, I have spoken to her before and would love
to hear it, but transformers have their issues too. Most of the wire
with gain designers steer clear of transformers for audio as among other
things they can mess up the fast transient low frequencies due to
saturation. Tube distortion is a flawed concept from 1950's designs. In
fact much of the distortion attributed to tubes is from the
transformers. Properly designed modern CCS biased tube circuits with
properly designed cathode follower buffers have as low distortion as
solid state and less distortion in the factors unmeasured. Solid state
is like sending music through a rock, tubes do not smear audio like
rocks because there is nothing in the way of hollow state. How does one
measure IM with actual music... When working with live instruments and
not smashing everything through compression the instantaneous peaks can
and do make solid state far more distorted. And it depends on circuit
topology as FET's produce more even harmonics than do tubes due to their
transfer function.
Step response does not provide the full picture but is a good starting
point. Sounds waves in real life are going in multiple directions
sinking and sourcing at the same time, direct and reflected phase
modulations etc the level of vector analysis would be crazy to know what
is really happening. I have heard amps with excellent step response
almost unmeasurable TIM and Harmonic distortion that sound like crap
because as you say all of the components to correlate and correct or
else some other reason. When the infinite network analyzer is built
then we will get much further. For now audio is still a somewhat black
art. When I find problems with a piece of gear someone has designed
that is often the comeback, "the impulse step response is as close to
perfect as possible in theory at room temperature". As I said before
plug a bass guitar in and that preamp that claims flat down two 2 Hz
suddenly sounds anemic and weak.
On 11/24/2018 9:08 AM, [email protected] [EVGRAY] wrote:
>
>
> Hi Mick,
>
> The current amplifier isn't ideal and as such has limited voltage
> swings. The output can't swing further than the supply voltage. A high
> Q inductor is closer to an ideal current source when disrupting the
> current through it. The voltage may reach high enough to punch through
> the insulation of the wire just to keep the current at the value it
> had during disconnection.
>
> The problem of amplifiers is also the use of lots of components in the
> signal path. Each component has its own resonance frequency because of
> not being ideal components. The feedback loop is for compensating for
> those errors. One solution to this is a wire with amplification or
> next to that an amplifier with a single stage or active component in
> the signal path like in the Zeus amplifier. This is like an old tube
> amplifier but with a modern transistor instead of the very distorting
> tubes (even though the distorted sound of tubes sound good in human ears).
>
> The Zeus amplifier: http://www.susan-parker.co.uk/zeus.htm
>
> "One of the biggest problems with audio designing is we cant test what
> is really going on wideband with music only guess or listen."
> Give it a step response by using a square wave as input. If not coming
> out as a sharp square wave the signal has changed. Just look at the
> oscilloscope probe for how to correct that. It costs a lot of
> amplification though.
>
> Regards
> Ole
>
>
> ---In [email protected], <mkjekyll@...> wrote :
>
> Ole,
>
> Yes the current buffer at the output end of the amp decauples the
> voltage amp from speaker reactance and is required and the more
> current the better all power amps have a basic currrent amp at the
> output. Most amp designs are hybrids. As you say with audio there are
> so many octaves and the impedance and reactance are non linear into a
> cored inductor, specially one that is a motor also reacting to air
> loading modes. A DC motor will speed up to destruction with unlimited
> voltage so what happens when the impedance goes super high or the
> speakers are not connected with an ideal pure CCS amp?
> Bass at the lowest impedance needs the most current. Highs like fast
> non slew rate limited voltage. Wrapping a design in loop feedback to
> keep it stable is really not good sounding and lazy practice but
> sometimes required. I use a curve tracer to match parts. Local
> degenerative feedback not an issue sonically but when it's wrapped
> around a whole circuit looped from out to in, yuck, at least have the
> individual parts nested.
>
> One of the biggest problems with audio designing is we cant test what
> is really going on wideband with music only guess or listen.
> Something that looks great with pink noise or impulse responsed can in
> practice not sound so good. Why many afficianados like SET designs
> because it's just one single ended tube without all the garbage, while
> not so good for rock and roll or stunning dynamics for less bombastic
> music the imaging can be superlative.
>
> For audio designs I really think Nelson Pass is solid, big respect
> here. In pro audio we like a shitload more watts on standby as every
> extra frequency divides power but for home gear can't go wrong with
> Nelsons designs and they are cheap and fun to build.
> http://www.firstwatt.com/faq.html
>
> Also a really good sound resource is Rod Elliot
> http://sound.whsites.net/index2.html
>
> This article goes into some of what we are talking about:
> http://sound.whsites.net/project56.htm
>
>
> On 11/24/2018 4:18 AM, onielsen@... <mailto:onielsen@...> [EVGRAY] wrote:
>
>>
>>
>> Hi Mick,
>>
>> The person here: http://nov79.com/amr/cont.htm claims that
>> current amplifiers are the best way to drive speakers. Citing
>> from page two:
>> "The current amplifier drives the speakers. It has no effect upon
>> the signal. It is the perfect black box which allows the voltage
>> amplifier to drive a heavy load as if it were no load at all."
>>
>> The only problem is for mass production as the two PNP
>> transistors and the two NPN transistors must be matched. For low
>> signal the solution is buying the transistors as IC packages as
>> they're matched by using the same crystal for all four
>> transistors. The person describes three sizes of amplifiers.
>>
>> This could be used for making a cheap differential probe or
>> current probe amplifier for scopes. The conventional ones are
>> pretty expensive to buy.
>>
>> "Audio is a weird duck as to an RF engineer it looks like but a
>> small notch however there are some really big differentials in
>> that small window concerning how materials react likely having to
>> do with the long wavelengths and the (phase?) change around 10kHz
>> where radio emission begins."
>> To cover the hearing range of human ears the amplifier has to
>> cover around seven octaves. That's pretty much. A human eye sees
>> only one octave. A RF transmitter may be even more limited.
>>
>> Regards
>> Ole
>>
>>
>>
>> ---In [email protected] <mailto:[email protected]>,
>> <mkjekyll@...> <mailto:mkjekyll@...> wrote :
>>
>> Ole,
>>
>> Current amplifiers are not a really a good choice to drive
>> speakers anyways as high impedance will get more and more voltage
>> with a CCS. Better to use voltage amps biased with current
>> sources. Dynamic speakers present a varying impedance bandwidth
>> so tweeters will begin to clip, unless the speakers are
>> electrostatic where the impedance rises as freq drops inverse to
>> dynamics. Current amps are a good choice for some preamp
>> circuits though which are driving a less reactive much more
>> constant impedance and as you claim can then leverage the
>> bandwidth of transistors. I am having an issue with a current
>> pre design right now that has huge bandwidth on the bench. On a
>> scope impulse response is flat from several Hz to almost a mHz
>> however when used for recording real world signals rather than a
>> sweep the low end is much thinner sounding than it should be. I
>> will try a diff power supply to see if massive current reserves
>> will make a difference.
>>
>> Thanks for the probe link, I will read in a bit.
>>
>> Interesting enough that audio is normally considered to low freq
>> for skin effect to matter as most RF engineers will attest but
>> silver plated oxygen free copper makes a big difference in sound
>> quality. The old fashioned wire wrap is often very good for
>> audio as it is silver plated copper and very thin and very
>> inexpensive. Good way to make interconnects that rival those
>> expensive ones for a few bucks. Coax is not really good for
>> audio either as presents too many non linearities. Audio is a
>> weird duck as to an RF engineer it looks like but a small notch
>> however there are some really big differentials in that small
>> window concerning how materials react likely having to do with
>> the long wavelengths and the (phase?) change around 10kHz where
>> radio emission begins. I imagine you can explain this much
>> better than I just did and I don't know the proper terminology of
>> where AF starts becoming RF which is much lower then the formal
>> claim of about 100kHz.
>>
>>
>> On 11/23/2018 6:27 PM, onielsen@... <mailto:onielsen@...>
>> [EVGRAY] wrote:
>>
>>>
>>>
>>> Hi Mick,
>>>
>>> Here is an article about oscilloscope probes: 'The Secret
>>> World of Oscilloscope Probes'
>>> (http://www.dfad.com.au/links/THE%20SECRET%20WORLD%20OF%20PROBES%20OCt09.pdf).
>>> It's not only the compensation network that matters. For
>>> minimum reflections the damping has to be done along the
>>> length of the cable. Just look at those thin coaxial cables
>>> on scope probes. I guess they would sound good as audio
>>> signal cables if used with the compensation network correct
>>> adjusted. It does require some extra amplification to
>>> compensate for the damping.
>>>
>>> Regards
>>> Ole
>>>
>>>
>>>
>>> ---In [email protected]
>>> <mailto:[email protected]>, <mkjekyll@...>
>>> <mailto:mkjekyll@...> wrote :
>>>
>>> Ole,
>>>
>>> Looks like feed forward compensation, sorry wish I could
>>> read Danish but looks like something fun to try. Current
>>> amps have high bandwidth but in my experience more dc offset
>>> from the common emitter input.
>>>
>>> On 11/23/2018 1:33 PM, onielsen@... <mailto:onielsen@...>
>>> [EVGRAY] wrote:
>>>
>>>>
>>>>
>>>> Hi Mick,
>>>>
>>>> The compensation network sits at the input end of the
>>>> amplifier. As for oscilloscope probe compensations they
>>>> can be placed at either end of the probe cable. As long
>>>> as the output signal of the speaker is an amplified
>>>> copy of the input signal at the source the compensation
>>>> does its job fine. For an audio amplifier the signal at
>>>> the speaker terminals should be an amplified copy of
>>>> the signal at the source. If placing the compensation
>>>> at the output end a lot of power is wasted. The
>>>> compensation has its own battery powered amplifier to
>>>> keep the signal level unchanged. The principle is shown
>>>> in figure 4 here:
>>>> https://ipaper.ipapercms.dk/TechMedia/AktuelElektronik/2007/13/?page=6.
>>>> The compensation (roll-off correction in the article)
>>>> is inserted in the signal chain as shown in figure 3.
>>>> Figure 2 shows the signals without the compensation.
>>>> His version is switched in steps while oscilloscope
>>>> probes use a variable capacitor.
>>>>
>>>> It is told that by putting resistance in series with
>>>> the speaker the sound is improved. This makes a voltage
>>>> amplifier act more like a current amplifier.
>>>>
>>>> There is an old B&O speaker with high damping that
>>>> sounds very good but wastes a lot of power. It requires
>>>> quite much power just to operate that model at moderate
>>>> level because of its high damping. I don't know if it
>>>> is damped by resistance or if it is mechanical damped
>>>> or both. A good way of driving a speaker is with a
>>>> current amplifier like used in scopes for their high
>>>> bandwidth. Of course the output current must be greater
>>>> for driving a speaker. Current amplifiers can be made
>>>> quite simple if using matched pairs of transistors and
>>>> the bandwidth can be something like 40MHZ for standard
>>>> transistors. This is fast enough to give a very good
>>>> replication of the finer details of an audio signal.
>>>>
>>>> Regards
>>>> Ole
>>>>
>>>>
>>>>
>>>> ---In [email protected]
>>>> <mailto:[email protected]>, <mkjekyll@...>
>>>> <mailto:mkjekyll@...> wrote :
>>>>
>>>> Hi Ole,
>>>>
>>>> Same concept we use when designing a zobel network for
>>>> a transformer in a mic preamp. I suppose this magic
>>>> device sits near the speaker. I have a hard time
>>>> believing this would do much with most dynamic speakers
>>>> as the mechanical impedance and reactance makes a
>>>> square wave hard to recognize. On an electrostatic
>>>> speaker though one can see a pretty good square wave
>>>> however that is one huge capacitor compared to some
>>>> trimmer so likely would require a tunable inductor in
>>>> line with the step up transformer. Of course I am
>>>> referring to using a laboratory mic for looking at the
>>>> square wave not just a probe across the speaker terminals.
>>>>
>>>>
>>>> On 11/22/2018 11:29 AM, onielsen@...
>>>> <mailto:onielsen@...> [EVGRAY] wrote:
>>>>
>>>>>
>>>>>
>>>>> Hi Andreas,
>>>>>
>>>>> Yahoo doesn't like to transfer the complete
>>>>> messages today even though the delay isn't too
>>>>> great. I completed the message history from the
>>>>> picture and hope it gets through.
>>>>>
>>>>> To make your own BNC feed through get one male and
>>>>> one female 50 Ohm connector and an inductance free
>>>>> 50 Ohm resistor. Then connect them all in
>>>>> parallel. Instead of one 50 Ohm resistor perhaps
>>>>> use two100 Ohm resistors in parallel. The
>>>>> resistors must be able to dissipate the power at
>>>>> the max voltage across them. With a low impedance
>>>>> current shunt this voltage won't be great. The 50
>>>>> Ohm feed through must be placed as close as
>>>>> possible to the input of the scope's input
>>>>> amplifier which means the BNC input terminal of
>>>>> the scope.
>>>>>
>>>>> For high speed oscilloscope probes the inner
>>>>> conductor of the coax cables are made of resistive
>>>>> wire to dissipate energy along the wire. This
>>>>> reduces reflections when the signal reaches the
>>>>> ends of the cable.
>>>>>
>>>>> I know of a compensation device like used in
>>>>> oscilloscope probes (i.e. like 1:10, 1:20, 1:50,
>>>>> 1:100 etc.). But this device is for audio signals.
>>>>> He tells that when it's correct adjusted the sound
>>>>> picture has depth end everything placed correct.
>>>>> He got the idea from oscilloscope probes where you
>>>>> want to see the signal on the screen like how it
>>>>> is at the probe tip. Never mind how it looks along
>>>>> the signal path as long as the end result is
>>>>> correct. That's what the compensation is for. I.e.
>>>>> when the compensation is correct adjusted an input
>>>>> square wave looks like a square wave on the
>>>>> screen. Without the compensation the corners of
>>>>> the square wave are round. With too much
>>>>> compensation the corners becomes pin shaped.
>>>>>
>>>>> Regards
>>>>> Ole
>>>>>
>>>>>
>>>>>
>>>>> ---In [email protected]
>>>>> <mailto:[email protected]>, <freeen2012@...>
>>>>> <mailto:freeen2012@...> wrote :
>>>>>
>>>>> Hi Ole,
>>>>>
>>>>>> On 22 Nov 2018, at 18:51, onielsen@...
>>>>>> <mailto:onielsen@...> [EVGRAY]
>>>>>> <[email protected]
>>>>>> <mailto:[email protected]>> wrote:
>>>>>>
>>>>>> Both ways can be used. For low inductance of
>>>>>> a current shunt let the return wire go close
>>>>>> to and in parallel with the shunt resistor to
>>>>>> minimize any area for the magnetic field.
>>>>>>
>>>>>> Here the return wire is twisted around the
>>>>>> current shunt.
>>>>>>
>>>>>> Also the 50 Ohm feed through is necessary.
>>>>>> Without it a lot of reflections will happen
>>>>>> unless the scope itself has a 50 Ohm input.
>>>>>> As scope inputs usually are 1M Ohm the signal
>>>>>> w ill be reflected due to the great impedance
>>>>>> mismatch from 50m Ohm to 1M Ohm and the
>>>>>> signal will drown in noise. 50m Ohm to 50 Ohm
>>>>>> reduces the reflections a lot.
>>>>>>
>>>>>> Thanks for the pic (illustrates it well). I
>>>>>> already have such a 50 Ohm feed through but
>>>>>> never really understood what exactly it does
>>>>>> and therefore have never used it. I
>>>>>> understand it has an impedance of 50 Ohm like
>>>>>> the coax cable to prevent reflexions, yes,
>>>>>> but how does it transfer the potential
>>>>>> difference to be measured (over the shunt) to
>>>>>> the 1M Ohm impedance input of the scope. If
>>>>>> we had to build such a 50 Ohm feed through
>>>>>> what would we have to do (what is it composed
>>>>>> of)? This might help me to understand how it
>>>>>> works.
>>>>>>
>>>>>> Thanks a lot,
>>>>>>
>>>>>> Andreas
>>>>>>
>