YAP power amp revisited, now at v2.1

[stinius]

It’s just a link about current mode versus voltage mode.
Not meant to distract the thread syn08.

Current mode

 

[scott wurcer]

It’s just a link about current mode versus voltage mode.
Not meant to distract the thread syn08.

Current mode

This is the strangest technical paper I've seen in years. I will forward this to Barrie for comment.

 

[syn08]

This is the strangest technical paper I've seen in years. I will forward this to Barrie for comment.

On the same note, everything can be debated, e.g.

"Negative feedback circuits do not exist, negative feedback is only an analytical instrument sometimes useful in analyzing circuits".

True or false?

 

[scott wurcer]

On further reading the philosophical weight applied to voltage vs current feedback is inappropriate. Likening this to, for instance, the delayed choice experiment in quantum mechanics is simply pretentious BS.

 

[darkfenriz]

Aghhh...
I've read it. Are some herbs already legalized in (edit) Switzerland?

 

[stinius]

Aghhh...
I've read it. Are some herbs already legalized in (edit) Switzerland?

I just couldn’t resist linking to the paper.

 

[Wavebourn]

This is the strangest technical paper I've seen in years. I will forward this to Barrie for comment.

I'm not retired yet, otherwise I would not resist to write a similar philosophical paper about a power mode.

 

[PHEONIX]

THD simulation

Hello Syn08

How well does Cadence 16.2 ( I assume this is what you are using) simulate THD, how low THD wise does it go down to.

Regards
AR

 

[syn08]

Re: THD simulation

How well does Cadence 16.2 ( I assume this is what you are using) simulate THD, how low THD wise does it go down to.

As good (or bad) as any other Spice simulator. They all, more ore less, share the Fourier analysis module from the original Berkeley Spice release.

Just follow the same rules as for LTSpice THD simulation and you should be, FWIW, fine.

 

[PHEONIX]

THD simulation

Hello Syn08

Thanks for the reply . Interestingly in the past Spectre the best cadence spice simulator states as one of its key specs that it can accurately simulate THD to below 1ppm (-120dB) , I didnt know that cadence 16.2 was good to these levels and hence the reason for asking the question.

Regards
Arthur

 

[jan.didden]

Re: THD simulation

Hello Syn08

Thanks for the reply . Interestingly in the past Spectre the best cadence spice simulator states as one of its key specs that it can accurately simulate THD to below 1ppm (-120dB) , I didnt know that cadence 16.2 was good to these levels and hence the reason for asking the question.

Regards
Arthur

AFAIK Spectre uses the harmonic balance method to calculate THD which is different from what most spice simulators use. I don't know the details but harmonic balance is often used to calculate THD in RF circuits and apparently does a good job.

jd

 

[PHEONIX]

Thd

Hello JD

It apparently can also do continuous Fourier analysis as apposed to the FFT method used by most other simulators .

Regards
AR

 

[syn08]

Re: Re: THD simulation

AFAIK Spectre uses the harmonic balance method to calculate THD which is different from what most spice simulators use. I don't know the details but harmonic balance is often used to calculate THD in RF circuits and apparently does a good job.

SPECTRE is one of the few surviving code branches having little to do with the original Spice. It had (has?) a very limited success outside some niche applications (like RF).

 

[jan.didden]

Re: Re: Re: THD simulation

SPECTRE is one of the few surviving code branches having little to do with the original Spice. It had (has?) a very limited success outside some niche applications (like RF).

Yes. Paul Tuinenga, one of the original founders of PSpice, tried his hand on an RF simulator, I think it was called Avista or something like that. I understand it never really got off the ground.

jd

 

[PHEONIX]

Re: Re: Re: THD simulation

SPECTRE is one of the few surviving code branches having little to do with the original Spice. It had (has?) a very limited success outside some niche applications (like RF).

Hello Syn08

I thought Spectre was one of the most commonly used simulators for analog IC design because it has superior convergence performance and accuracy ( provided you have good models) from what I can gather.

Regards
Arthur

 

[scott wurcer]

Re: Re: Re: Re: THD simulation

Hello Syn08

I thought Spectre was one of the most commonly used simulators for analog IC design because it has superior convergence performance and accuracy ( provided you have good models) from what I can gather.

Regards
Arthur

It's special applications only AFAIK. A SPICE simulation can converge with the same models to -300dB (double precision FP) but the simulation time for a long transient analysis can be prohibitive. Consider the problem of sigma-delta converters running at 10MHz clock and 1kHz input, msec of simulation at nsec time steps.

 

[PHEONIX]

Convergence

Hello Scott

I know that spice simulators can get very good results on THD simulations but not all of them do it AFAIK , and not all of thems convergence perfomance is the same . Would you agree with this.

Regards
Arthur

 

[kl]

Hello syn08,
Thanks for answers.

I have tried a simulation but currents through Q104 and Q105 are too low: ~0.5mA and ~4mA. How much should be these currents (OPS input stage and drivers)? D102 and D117 are infrared, but I see that they are green on PCB photo.
What is the bias current per MOSFET pair?

Thank you

 

[scott wurcer]

Re: Convergence

Hello Scott

I know that spice simulators can get very good results on THD simulations but not all of them do it AFAIK , and not all of thems convergence perfomance is the same . Would you agree with this.

Regards
Arthur

This is true. You need to look at what you mean by convergence. I usually consider lack of convergence to be complete failure to find a solution, within stated accuracy. Better simulators can switch integration techniques on the fly to help out. The second issue with respect to THD is to force convergence at exact time points and pad your transient analysis to settle all initial transients.

A common mistake is to not realize that some stock SPICE transient analyses will just run keeping the outputs within the specified tolerances and interpolate in between arbitrary points for the requested plot data.

 

[syn08]

Hello syn08,
Thanks for answers.

I have tried a simulation but currents through Q104 and Q105 are too low: ~0.5mA and ~4mA. How much should be these currents (OPS input stage and drivers)? D102 and D117 are infrared, but I see that they are green on PCB photo.
What is the bias current per MOSFET pair?

The LED model should have a voltage drop of about 1.8-1.9V as any standard green LED. I customized a diode model to accomodate this:

.MODEL GLED D
+ IS=10.000E-9
+ N=1.9388
+ RS=1.2366
+ IKF=45.752E-15
+ CJO=25.000E-12
+ M=.3333
+ VJ=.75
+ ISR=10.010E-21
+ BV=5.4169
+ IBV=10
+ TT=5.0000E-9

This model should not be used for anything else but to model a standard green LED forward drop.

Current through the input pair should be around 1.1-1.2mA and the output devices should run at about 150mA each. The current mirrors and the diamond buffers run at about 10-12mA.

The front end stage and the global feedback loop can barely be modelled in the time domain, due to the poor opamp models. It can be modelled in AC, though, to analyze the global feedback loop gain.

You can though analyze the output stage frequency response, loop gain, the effect of various compensation techniques (Miller, TPC, TMC), distortions, headroom, and make and educated choice of your implementation.

I do professional PCBs with ground planes, therefore I can go with the unity loop gain frequency up to 10MHz (YAP 2.1 has 4MHz, the original YAP had 8MHz). This may not be possible with home etched boards, so you may want to adjust your output stage unity loop gain frequency to 1-1.5MHz (and the unity global loop gain frequency to 300-400KHz) which is much easier achievable. You will sacrifice some performance though, but still be in the 10ppm THD range.