Modifying QPPM to measure "Max. True Peak Level"

[buildafriend]

The difference of "quasi" and "true" is a matter of attack time. A "quasi" takes a few ms to follow a fast peak, a "true" follows as fast as it gets. The latter gives a better indication of overload (important for ADCs/DACs), the former is perhaps visually a bit more pleasing.

Samuel

... interesting.

 

[bruce0]

And my whole desire is, to convert a QPPM into a TruePPM
...
Thanks for the answers!

Brgds,
Alex

Alex:

Are you trying to

a) Convert a physical piece of equipment? ( Is it the NTP Peakmeter model 177-400?)
b) Understand the difference between TPPM and QPPM?

If it is "a", let us know what you are trying to hack.
If it is "b", I think I am with Samuel Groner, in that the differences between these metering schemes (though I could not find a formal definition of QPPM) is about time and not level. 

Please forgive the following if it is obvious, but here is the way I would explain it:

If you consider an audio signal with a component of that signal being sine wave of lets say 20K HZ running 2V peak to peak, then if you measure a level of that for of 1/40,000 second you are going to see, and rectify it. You might see 1V.  But you also might see something a lot less than that, depending upon what phase component of the signal you sampled.

(This is just 1 example of why "instantaneous" Peak does not mean anything.  Measure in an instant, and you could measure 0 for a big signal.)

If you sample slower...( well there it is... Time  and that should take you back to Samuel Groners comment on attack time).

If you sample slower, say 3 milliseconds you will get 60 peaks and valleys of a 20K HZ signal, but possibly miss entirely measuring a 100Hz signal (which is very audible, very powerful, and more likely to overload things or clip audibly than the high frequencies are). ( You could still measure near 0 when there is a big 100Hz signal around)

These examples might make clear why a meter with a fast attack and a slow decay, makes sense of this type of signal.  The fast attack lets it respond to a high frequency overload, and a slow decay keeps the peaks of the lower frequency history on hand to display. 

I read your comments and questions about TPPM and QPPM as if they are standards that can be compared and converted, when it seems to me that they are just two metering schemes/algorithms , which measure level OVER TIME and are very dependent on the algorithm AND THE LOAD (which is to say the input signal, and it's history... whether a 3ms history or a 200MS history) to come up with a reading.

As to why the two schemes measure a sine wave the same way... that is probably because the sine wave is not varying over time and it's frequency is faster than the decay time of both QPPM and TPPM meters


So I guess my questions stands as:

What are you trying to hack, or what are you trying to understand, or just exactly what are we trying to establish here?

Warm regards,

bb

 

[electro_aLex]

I know already the difference between TPPM and QPPM and I am NOT with Samuel Groner ;-) !

For the definition of QPPM: http://www2.ak.tu-berlin.de/~akgroup/ak_pub/2008/Weinzierl_2008_Handbuch_Weinzierl_K10_Aufnahmeverfahren.pdf
- page 556
I am sorry for the fact that I cant find any clear, english literature about the topic, but google translater is your friend!

And yes, I want to convert a physical equipment (Analog QPPM) for measuring "True Peak" after the defination of "Max. True Peak Level" inside the EBU R128 Recommendation (0ms integration time ("fast" mode of one of my NTPs = 0ms!).

The only thing I dont know is: WHAT exactly do I have to hack?
Do I have to brigde a part?
Do I have to adjust the gain to "x1"?

Thank you guys! ;-)

Brgds,
Alex

 

[abbey road d enfer]

And yes, I want to convert a physical equipment (Analog QPPM) for measuring "True Peak" after the defination of "Max. True Peak Level" inside the EBU R128 Recommendation (0ms integration time ("fast" mode of one of my NTPs = 0ms!).

If you already have that, which is the essential part of having true peak detection, then you only have to correct the scale factor AND modify the scale to have 100%fs at the top.  Now I haven't the time (nor the curiosity) to check the EBU R128 norm in details, but you may well be faced with the fact that the standard has been conceived with a specific application in mind, that would probably include some idiosyncracies, in particular if it's intended for broadcast.

The only thing I dont know is: WHAT exactly do I have to hack?

Someone may answer if you published the schemo.

 

[electro_aLex]

I have a NTP 477, but I dont have a schematic, so I decided to publish everything I found in the net (277-400/ schematic of 177-400). I thought that this could help to identify the right spot in the schematic.. and because they all fullfill the same IEC/DIN norms, and its just analog electronics, hopefully the 177-circuit is representative enough to find out what to hack.

 

[audiomixer]

Alex,

I believe you have a misconception of the 'new' standard.... this is about having 'equal' loudness between different programs and uses digital metering including different long and short term integration to provide information about the loudness (perceived volume) of audio (or better TV) material. this is quite independent from metering before AD converters or level in a analog chain..... the different implementations of QPPM as per your pdf show just one thing - different tastes across the different broadcast houses around the world have grown different integration times, different headroom needs and the like, much to the discomfort of everybody in my opinion..... but the loudness level has been shifted over the years due to hopeless compression (while retaining the average QPPM level). so the new stadard measures something completely different.

now to peak metering in the analog domain: you cannot really see peaks  (thinking of digital clip) as they are almost infinitively short term (to the eye at least) and therefore peak meters might have a peak hold led that shows clipping, but then again there might be inter sample clip (see TC electronic white paper an that, even more confusing, but still true)....

a) I dont think you can mod a QPPM meter to 'digital kind' peak meter - you would miss the clip occuring...

b) the new level metering are far more complex as they meter average program level (among other). this requires integration over different time base and is likely to be best done in the digital domain.

c) the 0.707 factor is a linear gain thing and adjusts the 0dB reference point, nothing more.

d) you seem to be in a theoretical debate with your friend, not into modding real equipment (for Bruce0)....



just my thoughts,

- michael

 

[electro_aLex]

Alex,

I believe you have a misconception of the 'new' standard.... this is about having 'equal' loudness between different programs and uses digital metering including different long and short term integration to provide information about the loudness (perceived volume) of audio (or better TV) material. this is quite independent from metering before AD converters or level in a analog chain..... the different implementations of QPPM as per your pdf show just one thing - different tastes across the different broadcast houses around the world have grown different integration times, different headroom needs and the like, much to the discomfort of everybody in my opinion..... but the loudness level has been shifted over the years due to hopeless compression (while retaining the average QPPM level). so the new stadard measures something completely different.

now to peak metering in the analog domain: you cannot really see peaks  (thinking of digital clip) as they are almost infinitively short term (to the eye at least) and therefore peak meters might have a peak hold led that shows clipping, but then again there might be inter sample clip (see TC electronic white paper an that, even more confusing, but still true)....

a) I dont think you can mod a QPPM meter to 'digital kind' peak meter - you would miss the clip occuring...

b) the new level metering are far more complex as they meter average program level (among other). this requires integration over different time base and is likely to be best done in the digital domain.

c) the 0.707 factor is a linear gain thing and adjusts the 0dB reference point, nothing more.

d) you seem to be in a theoretical debate with your friend, not into modding real equipment (for Bruce0)....



just my thoughts,

- michael

Youre totally wrong! Please really read page 556.
Surely I can see the peaks, because the EBU norm says 0ms for the "True Peak" measuring, and this is slow enough. Why do you think I missunderstood the norm? I always wanted to meaure peaks, not loudness. Please read the whole thread, before answering such stuff....

What I need is just someone whos good in analysing the NTP 177 schematic, someone whos reallly "into" analog electronics.
I did enough research to explain exactly what I want.

We dont have to waste time with debating about things that are clear already.

 

[abbey road d enfer]

I have a NTP 477, but I dont have a schematic, so I decided to publish everything

Where? Again, if you're sure that your unit has zero attack time, you should have only to change the reference level.
The algorithm described in ITU-R BS 1770 is pertinent to evaluating the intersample peaks that occur as a result of the transient response of the antialiasing filters in A/D converters. These are pertinent only to digital measurement. In analog, there are no intersample peaks, because analog is a continuous process. Basically, all you need is an absolute value rectifier followed by a 1st-order hold circuit, with an attack time << than the shortest transient (think 17.5uS for 20kHz BW).

 

[JohnRoberts]

It's poor form to get angry with the people who are trying to help you.
-----
It isn't simple to change just one aspect of a circuit without affecting something else.

It looks like c10 is both controlling the attack and release time constants, so making the C smaller will also screw up the release time that you still want to be slow.

The charging attack seems to involve several active and passive parts Q7,Q8,Q9(?) and R18,19,20.  Since the schematic includes a table showing values for DIN and Nordic standards. 

I'd experiment with making the resistor values smaller but wouldn't expect a night and day difference between PPM and faster, and there looks like some level dependent fast attack tricks involving the transistors and perhaps part of the log conversion to get a dB scale, (while I am not familiar with the drawing convention they use for transistors).

JR

[edit] disregard... my comments were in the context of a schematic posted for a different analog version. [/edit]

 

[audiomixer]

why do you think I did not read the thread, but hey, don't mind.....

You do need to integrate your signal on a analog meter, viewing one led light up at 44.1khz (0.02ms!) for one clipped sample is just not possible. your digital meter (read audio workstation) can display the peak value as a numerical info (like -0.1 dB) but it's bargraph has an integration time too...  eal world analog meters can't be infinitely fast.  the peak metering integration time is defined as below 1ms - that is not 0ms, think more 0.5 to 0.90ms.... if it were faster say an order of magnitude your typical specs would say less then 0.1ms

there is someone posting on a german forum with a NTP diagram... see here:
http://www.mikrocontroller.net/topic/277375  with a remark regarding the rms conversion in the schematic, but this is wrong the rms conversion is performed in the full wave rectifier right up front. the rectified signal is then integrated to match the required ballistics - thats where you want to change the 'type' of metering from QPPM to almost true peak. the log converter is to display the metered value in dB, so that you have a sensible display. there are some more bells and whistles, but this is it basically.

the NTP 477 range is digital with dsp chip (as per NTP brochure) - but this again is a completely different meter, more like what you find in a workstation, or uwe beis FPGA implemented super duper meter.
http://www.beis.de/Elektronik/DPLCM/DPLCM.html he shows peak value in numerical form too.

NTP info on the meters here: http://www.guano-grafik.dk/pdf/ntpkatalog.pdf
you got sample peak (with hold of course in there already


open up your meter and post a pic of the big TI chip ;-))

cheers, Michael

 

[abbey road d enfer]

Well, I have checked the schemo for the 177. the attack time seems to be governed essentially by the dynamic resistance of Q7 and Q9 against capacitor C10. Since it happens in the log domain, it's not an easy calculation.
But it's definitely not zero-attack. Simulation gives ca. 5mS, but accuracy is about -80/+400%.
I can't imagine any simple mod to reduce the atk time enough to make it zero in the context of 20kHz BW.
This architecture is just not convenient.

 

[electro_aLex]

Once again: I can set the attack time to 0ms on my NTP, so there isnt any hack for the attack/integration time needed! And definitely zero is of course way too fast to see...

Just this 0,707 gain setting has to be changed to 1!

Brgds,
Alex

 

[abbey road d enfer]

OK, so it seems you have a digital peak-meter.
It would have been much easier if you said that earlier, could have avoided a number of false tracks.
Now, HOW do you want to hack it? Hammer and chisel or software?
The "zero" atk time is done in software - or not, but only the manufacturer (or the technical manual) can confirm if the intersample detection is buit-in. The peak-factor correction is a trivial subject compared to that.

 

[electro_aLex]

No mine isnt a digital PPM, its for analog audio, nothing with intersamples.
I thought of "hammer and chisel-hack" .
If its so trivial, please give a hint.

 

[abbey road d enfer]

Couldn't you say what your effing PPM is? That would have helped a lot.

 

[electro_aLex]

NTP 477-700.

 

[audiomixer]

Hi Alex,

please post a photo of the unit's pcb.... the NTP handbook says DSP processing for the 477-700 its just analog input...

Quote from the NTP leaflet, Page 8: "All signal processing is done in the
digital domain, with use of a Sigma
Delta converter and a DSP circuit to
achieve all the stability and accuracy
improvements associated with the
digital technique."

by the way intersample peak detection is quite new, something like 5 years-ish, before the TC paper on th subject this was unknown, and even they needed a few years to gain acceptance on the matter. but to be honest ISP (InterSamplePeak) is for the hardcore compressed lunatics that squish the last 0.05 dB out of audio material..... just leave a ever so slight margin and you will be fine.....

so nothing to do with the 177 / 277 lines besides the form factor ;-(

- cheers, michael

 

[JohnRoberts]

What a long strange journey this has been...

from the 477-700 literature. "The integration time is switchable (10ms/“zero“), to meet the require- ments of digital recording. One LED on the front plate clearly indicates the selected integration time."

Just set the integration time to zero and you are fast-peak.

JR

 

[abbey road d enfer]

NTP 477-700.

It's a digital meter! Unless you have a complete development system and the source code for whatever DSP is used in the unit, I don't see how you can implement the algorithms for intersample peak detection.

 

[electro_aLex]

Just digital signal processing Theres still the problem of eliminating the RMS-factor!
I cant take any pictures of my PPM at the moment, and I think if the signal processing is done digitally, hacking it is not easy.
But a friend of mine has a 277-400, maybe this device works totally with analog-circuits.
If I can identify some trim-potentiometers, how can I change the setting to get 29% more on the bargraph?
Do I just have to feed it with a 1Khz sine, and calibrate?

"Long strange journey" is maybe your opinion - if there were some more technicians arround, and if there hadnt been the same questions again and again, this journey wouldnt have been so long.

I deliberatly asked in a very general way, because I searched for a more general answer that can be referd to the most peak meters. Of course I understand that this wish cant be combined on the one hand with totally analog devices and on the other hand with semi-digital processing.

My opinion is, that eventually there are some people arround, who are really interested in this topic, because the advantage of this application would be, to measure true peaks, if we combine the "TPPM" with a DAC. In my backround of usage, this "TPPM" is just for fun, to see how it would perform.