Qustion about Peavy VYPYR-15 PA topology

[sajy_ho]

Hi guys, I have a question about TDA chip in Vupyr-15 power amp section:
http://music-electronics-forum.com/attachments/5037d1239065731-pv-vypyr-15-pa.pdf
What is the purpose of the current sense circuit in the speaker feedback loop? I read somewhere that it's there to raise amp's output impedance to mimic  tube amps; is that true? and how does it affect the sound?
Another question is how does it affect the amp's gain?

Thanks
Sajad

[teemuk]

The resistor develops a voltage drop that is dependent on load current. (I=U/R) Since reactive load, like loudspeaker, is not purely resistive but has impedance (IOW load resistance varies with frequency) the voltage drop across the "current sampling" resistor also varies with frequency.

This consequently varies magnitude of additional (negative) voltage feedback taken from that point. This effects closed loop voltage gain of the amp.

Dynamic loudspeaker has some archetypical impedance characteristics: Due to voice coil's inductance the impedance increases towards high frequencies. In addition, at resonant frequency of the speaker the speaker diaghram moves with greatest efficiency, which means the resonant frequency introduces a very high-impedance load in comparison to nominal load impedance. If nominal load impedance is, say, 8 or 4 ohms the impedance at resonant frequency and at upper high frequencies can be close to 100 ohms.

Google speaker impedance. Should be an eye opener.

You can now imagine a resistive voltage divider formed by the load impedance and current sampling resistor. When load impedance is moderately high (e.g. towards high frequencies and at resonant frequency) the resistive divider attenuates the additional feedback signal significantly and consequently voltage gain of the amp increases. When load impedance is low-ish (e.g. nominal impedance) the divider attenuates the additional voltage feedback signal less and closed loop gain decreases.

Thus the amp has higher voltage gain at higher load impedances. This shapes the frequency response: Gain is boosted at resonant frequency and towards higher frequencies.

Resonant frequency is typically near low end roll off of the driver while inductance increases towards upper high frequencies. Therefore result is boost of bass and treble, or in different terms: a mid-range notch.

If you think about an amplifier circuit with high output impedance (archetypal tube power amp) driving a reactive load it creates a similar frequency response due to similar divider formed by output Z and load Z. Simply, if output Z is high more voltage is developed across terminals of higher ohmic loads than across terminals of lower ohmic loads.

On the other hand, if output Z is very low (e.g. generic solid-state power amps) the divider has very little effect on voltage potential created across load terminals. This is great for high-fidelity since amp's frequency response is basically unaffected by load impedance. One acquires flat response despite impedance characteristics of the load. Very nice characteristic if additional "coloration" of the output signal is a no no.

With musical instrument amps (particularly guitar amps), however, we have grown to be accustomed to low and high frequency boost of a high output Z amp. So it's basically just another equalising circuit in the signal path. Amps without it sound lacking in bass and treble because they have flat response instead of boosting highs and lows.

This is a graph from whitepaper published by Fender's design engineers in early 1980's, which presents the difference of frequency response of high output Z amp (tube amp) and low output Z amp (solid-state) driving a reactive speaker load. In high output Z amp the impedance characteristics reflect in overall frequency response, in low Z amp the response remains moderately flat despite changes in load impedance.

Naturally, most SS guitar amp designers have been aware of this since late 1960's and have designed their SS amps to replicate the characteristic where frequency response is dependent on load impedance. For "Hi-Fi" (and largely for bass guitar amplification) such excessive coloration, however, is somewhat a "detrimental" characteristic. You mainly see this in guitar amps only.

So yes, it is true that this kind of circuit mimics similar behavior of tube (power) amps (but only those those that have high output Z) and likewise affects (closed loop) gain of the amp creating an impedance-dependent frequency response.

[sajy_ho]

Thank you so much Teemuk for your excelent explanation.


I'm planning to use this schematic for my DIY combo, and I'm wondering if I could lower the closed loop gain of the amp by changing that 39k resistor in the feedback loop to around 22k, because I'm already using a hi gain preamp and with that closed loop gain in the schematic; the amp goes into PA distortion very easily.

So can I change the 39k resistor to 22k without the danger of low gain unstability?

Thanks again...

[sajy_ho]

I built it yesterday according to schematic except changed R7 and R8 to 22k; now the amp sounds good but it gets so hot even with a oversized heat sink and at very low volume. Can changing R8 to 22k caused the overheating?

[J M Fahey]

Besides the "trivial solution" of the heat sink not being as large as you think  , the chipamp might be oscillating.

Since they are high gain very wide bandwidth devices they are picky about grounding, shielding and bypassing.

Notice ceramic decoupling capacitors C12/C14 , going straight from power pins 3 and 5 to nearby ground.
And I mean *nearby* , think 1 inch away or less.
Also Peavey designer himself wrote in big bold letters about Ground trace in the PCB: "LARGE SHORT TRACE" .
That´s unusual in a schematic so it must have been a clear message to Peavey´s own PCB designer  :trouble
R2 C18 stability Zobel network must also connect from TDA speaker out pin to nearby ground.

Another issue is that these chipamps need a certain minimum gain (usually > 20X to 30X) or they self oscillate.

Datasheet example usually suggests NFB resistors 22k-680 ohms, which would make gain= (22/.68)+1= 33X .

Peavey uses 39k-470 ohms which if used alone would give you 84X , quite acceptable, BUT you also have gain reduction by the current sensing network: 4 ohm speaker / 0.2 ohm resistor, some 20X .... quite on the "dangerous" side.

Maybe the Pevey amp *just*  worked with R8=32k .... after all the designer was worried about stability, they must have been quite on the edge, and your lowering it to 22k was the straw that broke the camel´s back 

In principle I would set it back to 39k and if you have way too much gain, use an input attenuator.

You might add 100k-47k or 100k-22k between preamp and chipamp input.

Try it and post results.

[sajy_ho]

Thank you so much. I'll change it back to 39K and see what happens. So if I change the feedback R to 39k, should I change R7(resistor from non inverting input to ground) to 39K so both inputs get the same bias current to prevent DC offset and noise?
Also here is my pcb drawing; I'll appreciate it if you take a look:
https://www.dropbox.com/s/nycdo2jf0qgauz0/TDA2040%20Guitar%20Amp.pdf?dl=0
And here is my build:https://www.dropbox.com/s/b2j2z817dz8z1gu/20170701_191114.jpg?dl=0


EDIT: There is somethin I don't understand; why negative current feedback reduces the gain? PS I'm using 16 ohm guitar speaker!

[sajy_ho]

I changed R8 to 39k but it's still getting hot even at lowest volume! Could someone please give me a formula for gain of mixed feedback?
Thanks

[phatt]

I just looked at the picture of your build,, your heat sink needs to be larger. 8|
Having said that I've opened a lot of small amps that use the same/similar power chip and they have less than ideal heat sink so yes they run hot and would likely last longer if the heat sink was larger.

With no signal after 10~15 minutes the chip should stay at room temp if not there is a problem.
As *J M Fahey* has already mentioned it could be an oscillation above your ability to hear and hence heating up even at idle.

These small amp chips will reach their limit quite fast and early distortion is common
Phil.

[sajy_ho]

I just looked at the picture of your build,, your heat sink needs to be larger. 8|
Having said that I've opened a lot of small amps that use the same/similar power chip and they have less than ideal heat sink so yes they run hot and would likely last longer if the heat sink was larger.

With no signal after 10~15 minutes the chip should stay at room temp if not there is a problem.
As *J M Fahey* has already mentioned it could be an oscillation above your ability to hear and hence heating up even at idle.

These small amp chips will reach their limit quite fast and early distortion is common
Phil.

I checked and it gets gradually hot after 10 minutes of running with input connected to ground!

[sajy_ho]

I found it, it was the chip! I swapped it with another TDA2040, still getting hot! Then I used a TDA2050 that I found around and it got hot again! Then I took my last chance and used another 2050; it worked without getting too much hot. I remember once somebody told me this toys(that's what he said!) are all fake today!

Anyway thanks for all your help...

[sajy_ho]

[quote author=J M Fahey link=topic=4264.msg33999#msg33999

Peavey uses 39k-470 ohms which if used alone would give you 84X , quite acceptable, BUT you also have gain reduction by the current sensing network: 4 ohm speaker / 0.2 ohm resistor, some 20X .... quite on the "dangerous" side.
[/quote]
I still don't get this; why the gain reduces by the factor of (Output Z/R sense) ?
Assume using a 16 ohms speaker instead of 4 ohm; so the gain must be reduces even more according to Mr. Fahey, right?
but I think using a higher impedance means less feedback from speaker and consequently it means reducing the gain even less, not more! Am I making any sense to you?

[J M Fahey]

EDIT: There is somethin I don't understand; why negative current feedback reduces the gain? PS I'm using 16 ohm guitar speaker!

To drive amp you must match NFB signal.
You have two NFB voltages there, which are in series, one coming from resistive attenuator 39k/470 , call it "X", the other from [speaker impedance]/0.2 ohms , call it "Y"
So you need to match "X" voltage , which would be 84X as calculated before, PLUS "Y" voltage.
So if you have to match a higher voltage, you need a higher signal to do it, so net gain is lower.

In a simple example, suppose conventional NFB is 10k/1k , so gain is 10X
To get 10V RMS out you need 1V RMS in.  That´s the conventional way.

Now suppose you have a 4 ohm speaker and a 0.4 ohm resistor.
NFB *there* will be again 1V RMS if amp is putting out 10V RMS

Now connect the 1k resistor, not straight to ground as usual but to the Speaker/0.4 ohm junction.

Hey !!! 0.4 ohms is almost zero ohms!!! It´s nothing compared to 1k!!!! It´s practically the same as connecting straight to ground !!!!  Nothing should change!!!!
Think again: that junction is not ground at all, since it does not have ZERO volts on it (as true ground would) but 1V RMS ... which so is in series and summed to the standard 1V RMS.

Now you need 2 V RMS to drive that amp to 10V RMS out ... so now effective gain was reduced to 5X

That´s why I said earlier that even if going from 39k to 22k "looked" safe on its own ... in practice it might not  ... specially with Peavey sanctioned 4 ohm (nominal) speaker .

[sajy_ho]

EDIT: There is somethin I don't understand; why negative current feedback reduces the gain? PS I'm using 16 ohm guitar speaker!

To drive amp you must match NFB signal.
You have two NFB voltages there, which are in series, one coming from resistive attenuator 39k/470 , call it "X", the other from [speaker impedance]/0.2 ohms , call it "Y"
So you need to match "X" voltage , which would be 84X as calculated before, PLUS "Y" voltage.
So if you have to match a higher voltage, you need a higher signal to do it, so net gain is lower.

In a simple example, suppose conventional NFB is 10k/1k , so gain is 10X
To get 10V RMS out you need 1V RMS in.  That´s the conventional way.

Now suppose you have a 4 ohm speaker and a 0.4 ohm resistor.
NFB *there* will be again 1V RMS if amp is putting out 10V RMS

Now connect the 1k resistor, not straight to ground as usual but to the Speaker/0.4 ohm junction.

Hey !!! 0.4 ohms is almost zero ohms!!! It´s nothing compared to 1k!!!! It´s practically the same as connecting straight to ground !!!!  Nothing should change!!!!
Think again: that junction is not ground at all, since it does not have ZERO volts on it (as true ground would) but 1V RMS ... which so is in series and summed to the standard 1V RMS.

Now you need 2 V RMS to drive that amp to 10V RMS out ... so now effective gain was reduced to 5X

That´s why I said earlier that even if going from 39k to 22k "looked" safe on its own ... in practice it might not  ... specially with Peavey sanctioned 4 ohm (nominal) speaker .

Thank you so much, I get your point quitely. What I was mistaking about was that I tought gain will reduce exactly by this proportion (Load R/R sense). In the mixed feedback mode, the current feedback is diffinitely reducing the gain but not by that simple proportion. I need to finally read Teemuk's book I guess...

Thanks again for all your help.