Here the chrominance path in the NV-HD100.
I choose that schematic because it has no so integrated blocks, the NV-FS200 has a M52083 for all the chroma demodulation.
It seems to be at near integer frequency distance from the FM carrier to the downconverted chroma carrier, but in the high frequency side.
The JVC reference mentions something about Fo + 2Fdc
Maybe in this case is something Fo + nFdc being n an integer +- 1/2
The recent SPICE simulation of the luma filter for the PAL panasonic shows a cutoff notch filter around 8MHz (the dB scale is confusing, it depicts it as two notches, but it is a bandpass centered at Fo with asymmetric transition bands)
It might be a strong image of the sync pulses at 2Fo (tomorrow I will tune gnuradio to that frequency to hear what it sounds like. If there was an image of the sync pulses we can hear them) It is not sync pulses related, it is something else.
Maybe the NTSC one has this cutoff around 7MHz.
The Wikipedia article is right for ideal square waves with infinite harmonics, real square waves have not infinite harmonics since physical propagation/transmission/recording mediums have not infinite bandwidth.
Any waveform on time can be expressed as a sum of sine cosine waves:
By the third harmonic it start to look as an square wave.
Take a look a this video, for me it is a lot better than complementary to the wikipedia and textbooks for undestanding Fourier's work.
The same channel have a lot of other interesting videos on the subject. (Not so easy one to wrap the head around, I'm still trying to catch its full potential)
I found interesting the fact that constant complex multiplication changes phase (it rotates the signal on the complex plane).
I finally understood why they choose the complex numbers to work with instead of let say R², it is all about math notation convenience.
I like to call them stereo numbers now.
+ Reply to Thread
Results 331 to 344 of 344
Last edited by VideoMem; 14th Oct 2020 at 15:20.
The above example is using a 10MHz inline BNC LPF, but he has done another test with a 13MHz Mini-Circuits LPF with even better results. This most closely matches the 13.2MHz filter design of the Domesday Duplicator.
Last edited by Titan_91; 17th Oct 2020 at 15:11.
By the simulations of the Panasonic NV-FS200, PAL VTR active luminance filter spotted by oln, the zero gain point is at 8MHz, and then, it opens again leaving about 10% of the frequencies above it.
Here are better plots of the SP and LP mode filter bandwidth (linear scale):
An 8MHz LPF instead the default 5MHz one can improve the sharpness of the picture.
However there are some issues with the setup I see on the pictures.
Driving a moderate length coax/BNC directly from the RF test point is not the best impedance match, it can cause some oddness/noise/attenuation.
The head amplifier is intended to drive an unbalanced high impedance line over short distances (PCB traces), I guess it expects a 120/300 Ohms line, not the 75/50 Ohms coax.
(One or) A pair of RF balun transformers can solve the line driver / cable issue, however in the case of the added RGB amplifier/buffer near the testpoint this might not be necessary.
The balun will also help to break ground loop type noise coming from the PC to the player and vice-versa.
I mean something like this:
RF Testpoint -> (0.1~1µF polyester/NPO/ceramic if DC present) -> balun -> coax -> 8MHz LPF filter -> capture card ADC
RF capture point -> (0.1~1µF polyester/NPO/ceramic if DC present) -> balun -> coax -> amplifier type (A) (with a separate power supply) -> coax -> 8MHz LPF filter -> capture card ADC
RF capture point -> amplifier type (B) (shared supply with the head amp) -> balun -> coax -> 8MHz LPF filter -> capture card ADC
With amplifier (A) I mean one of the channels of the RGB amp.
The (B) one is the one transistor alike solution posted above.
I've done some progress this days. I successfully decoded luma and displayed it in gnuradio. Here is the screenshot:
There is no ld-decode code involved at all here, so, there is no Sync, nor deinterlacing, TBC or dropout compensation (yet).
I'm currently writting an OOT module for Sync and TBC, since I can't figure out easily how to do proper sync and variable resampling using the modules provided by default in gnuradio.
Here is what it starts to look.
Last edited by VideoMem; 15th Oct 2020 at 17:01.
Here's the data sheet illustrating the PCI bus interface. PCI is not ideal though, I wonder if this can be adapted for use with single lane (1x) PCI-E? Or even USB 3.0? This would require modification of the driver though. At the moment it's still easy enough to find older PCs and motherboards with PCI on eBay for cheap.
Last edited by Titan_91; 16th Oct 2020 at 13:27.
RF filter/eq is also interesting, it's something I've found very little documentation on. The book I have and techical manuals describe demodulation of the signal using a double limiter, integrator etc, but don't really say much about how the signal is filtered before that other than there usually being a FM agc. The filter setup must be rather crucial though, I don't have the NV-HD100, but I do have it's big brother the SVHS NV-HS1000, and it doesn't take much altering of the SP Q/LP Q trimpots for black or white (depending on direction) streaks to show up, especially on more worn/lower output tapes. If I move the lowpass filter up much in vhs-decode from what it is currently, which is much lower than 8 mhz I also get streaks, so maybe there is more to it we're not seeing here, or maybe it should be higher up and sharper.
As for capturing the signal, personally I used a Sony SLV-SE60. It looks like it already has a 0.01 uf dc blocking cap for the RF test point, I used an oscilloscope probe both with the cxadc card and domesday duplicator, I don't know if that could cause problem or if it's a reasonable way of doing things. The domesday duplicator was designed to tap rf from laserdisc players and is set up for 50 ohm impedance I think. (Also I found later that the VCR had some over-aged power supply capacitors, which I have fixed later, so I don't know if they had any impact on the captured signal, it's been a while since I did the captures.)
I'm tinkering with dropout compensation in vhs-decode now, you can see this dropout quite clearly on the rf waveform:
[Attachment 55530 - Click to enlarge]
Blue is rf, red is demodulated, and green is signal envelope.
Last edited by oln; 16th Oct 2020 at 11:35.
here. Not sure what VCR it was used in.
There are full datasheets for several of the Sanyo ones, like the mentioned LA71750 used in many late-model VCRs made by LG, Samsung, Daewoo, Funai etc, LA71598HM and others.
Also found some for two of the philips chips used in the NV-HD panasonic models, TDA9715 and TDA9725. (I wonder why they used philips chips and not panasonic's own.)
Has anyone been able to find out what impedence the PB FM test point on a typical VCR seems to be? For LaserDisc it's between 47 and 50 ohms. Zcooger has used 75 ohm coax with good results with both PB FM and Hi-Fi test points.
I think I didn't put this info anywhere but Discord - I used 50Ω coaxial cable (RG-58). I didn't check or tried to match the impedance yet between VCR and PC but I want to know how to do it so the reflections disappear.
Maybe it has a cutoff at 5MHz but a transition band around 3MHz.
Yes, I would have to look it in more detail too. I annotated it with questions, because I was in the search of the "comb filter" thing.
Comb filtering only on recording seems to make sense, because fixed delays with tape playback flutter/jitter will cause oddness (I think)
Another thing I observed on 'how to detect dropouts' (which is the first step on dropout arbitrage) is the fact the demodulated signal often goes to sync level when in dropout on the display area. It is expected to the signal stay between 0.3 and 1V and not going lower than the black level (0.3V) on the display area section.
I think it could be a way to flag a dropout.
Thinking, maybe that's the reason why they called it 'dropouts', as a reference to something that being drop from some height.
The double limiter thing is a plain rails cut/saturation, lets say, you pass the FM through an AGC that normalizes the signal amplitude to be between 0 and 1V (but it has a gain adjust response about a line time), then you demodulate the FM, then you put a rail block (that's how they call on gnuradio) to ignore all the spikes coming lower than 0V and higher than 1V if the signal goes crazy between hsync pulses.
Here what I see on the demodulated signal when it dropouts:
The .grc file:
VTR head RF signal test_PAL_SDL_plain_raw.zip
The names of the tabs are residual from previous tests, the only valid is the Luminance one
Overaged electrolytics are a plague on these machines, and the beta ones, specially late 80s machines.
If it smells like fish, you have bad caps. ( The electrolytic inside the cap has a Pyridine based compound, which smells fishy when leak )
Camcorders suffer for that problem too.
If the leak is not fixed in time, it corrodes the PCB. The machines start to leak when plugged in after been sitting 20 years or so, watch them.
This guy has a lot of videos on troubleshooting helical scan tape machines, from U-Matic to DAT.
99% of the time are bad electrolytics in some critical filtering (not everywhere, no need to recap them all)
This video brought me here. He also hacks the players a lot.
I started with the intentions of decoding the PCM-F1 adaptor, then I found it was already done
Beta/Umatic seems to record the signal in a similar way.
Let say, you measure it at no load, and measures 1Vpp
Then you load it with a 120 Ohm resistor and measures X
Then you deduce the internal source thevenin's resistance impedance.
<- this voltage source and Rth are internal to the head amplifier, the goal is to determine Rth +------/\/\/\/\--------+ Rf testpoint | Rth (V) | ---
The circuit I've mentioned for the luma filter buffers the signal coming from the head amplifier prior to sending it to the filter.
So I guess the amplifier might don't have the power to drive a coax directly without hassle, (or at least they take special care in that design to not to load it with anything)
PD: I think USB 2.0 / 3.0 is the way to go, but PCI express can be a good solution too
Ground loops are an issue too, besides the impedance matching, a transformer that separates the ground from the machine and the capture device / RGB amplifier could improve the S/N ratio.
Last edited by VideoMem; 17th Oct 2020 at 06:00.
What type of balun would you recommend installing between the test point and RGB amp, and how should I solder it in?
Something like this (isolator + impedance adaptor?):
And / or this (only impedance adaptor):
I'm not sure if they gonna work. I you ask me, I will experiment winding one on a small ferrite toroid core with magnet wire.
But that enters the electronic craftmanship area.
Not sure if it can be done with already made commercial parts. These ones are the nearest I found.
The second one is the standard 4:1 non ground isolated CCTV balun.
The first one seems to be a 1:1 isolator 50 Ohms to 50 Ohms, but I'm not sure.
A signal generator capable of 3~5 MHz and an oscilloscope can be used to determine the transformer ratio.
The impedance ratio is:
Zp/Zs = (Np/Ns)²
Where Np is the primary turns and Ns the secondary turns, these are direct proportional to the voltage ratios between primary and secondary.
Zp is the primary side impedance, and the Zs is the desired secondary impedance (50 Ohms)
Let say you have a 4:1 turns ratio balun, if you apply a 5MHz @ 1Vpp sinewave to the primary it might produce 250mVpp in the secondary
Then you turns ratio is 4:1 because:
1 / 0.25 = 4
Zp = Zs (Np/Ns)² => Zp = 50 Ohms * ( 4 / 1 )² => 50 * 16 = 800 Ohms
Then it adapts a 800 Ohms line to a 50 Ohms line
A 1:1 isolator is what I think is needed to break the ground loop issue, but it only will work well if the RF testpoint output impedance is already 50 Ohms
The signal at the input of the ADC needs to be 1Vpp (peak to peak), as is present in the testpoint.
A 4:1 will reduce the amplitude to 0.250Vpp, but it will match the impedance of the coax cable.
To use the 4:1 directly the signal must be amplified 4x by a transistor stage and a class B push pull driver amplifier prior to sending it to the impedance ratio matching balun.
And the common ground between primary and secondary present on that devices must be disrupted to prevent the ground loop issue.
For soldering it, place it near to the head amplifier as possible with a twisted pair input soldered to the testpoint (the 800 Ohms side), and then use the coax/BNC output as is for connecting the rest of the chain.
I'm assuming the input impedance of the card / RGB amplifier is 50 Ohms.
If the card already have a programmable (analog) gain frontend, then the amplifier stage prior to the balun is not needed.
Last edited by VideoMem; 17th Oct 2020 at 18:12.
Would a DC removal capacitor installed in series work for isolating ground as well?
Last edited by Titan_91; 19th Oct 2020 at 22:17.
It will decouple the low frequency component of the ground loop noise, but not the MF ~ HF band component.
The signal has picture / machine information from 50Hz to at least Fh * (284-1/4) + Fh/625 for PAL-B,
and 455 * Fh / 2 for NTSC, as I interpret from the JVC reference
Fh = 15.625 kHz (PALB) or 15.734 (NTSC)
It yields 4.43362 MHz for PAL and 3.57949 MHz for NTSC
The capacitive reactance of a capacitor of 10nF (0.01µF) at 1 MHz (yes, 1 MHz) is
XC = 1/(2piFC) = 1 / (2 x 3.141592653 x 1 MHz x 10 nF)
= 15.9 ohms (to 3 significant digits)
Which is comparable to the magnitude impedance of the transmission line.
A 100pF capacitor (estimated stray capacitance between primary and secondary of the balun) have a reactance of about a megohm and half at the same frequency.
I'm not expert on this particular subject, maybe someone who is skilled with the MF ~ HF band of "amateur radio" could help us with the best solution here.
Maybe shango066, MrCarlson from Mr Carlson's lab, Dave from EEVBlog, or CuriousMarc.
(in band frequency order )
That band is a mess with SMPS switching noise present in the modern PCs and home appliances.
Even a VCR with an unshielded RF case could incorporate noise to the signal.
Capacitive coupling between the VCR ground and the PC ground is far from ideal, not only for the signal, also for the VTR electronics.
Who could guarantee that the servo controller loop will perform as desired in the presence of a strong ground noise coming from the PCs and surrounding appliances.
These devices have resurrected in an strange atmosphere, that's wasn't present as strong in the days the specs were written.
Could be reflections possible at all with 10 meter wavelength in cables up to 1 meter length?
RF is funky.
Some measurements should be conducted with real equipment.
Maybe besides the picture information specs there is vestigial recording of other things in some parts of the signal.
Teletext as an example beween fields,
NICAM audio? I don't know anything about NICAM audio yet.
A2 Zweikanalton? What?
Does a capture here have macrovision enabled?
I recently saw a resume of the domesday86 project here:
The whole concept were erased from my head.
I refreshed it. How far we are from that milestone point?
I'm not from UK, USA, Canada, France, Japan or Australia, and I'm far from having the proper equipment to measure things, I don't have access to the machines.
You have them.
Even the internet here is not what it should be for 2020.
I will inspect the captures/write software as far as I can, but I can't experiment with anything physical, the specific knowledge I have from the subject comes from the setting up of experiments, internet, and the introduction from the formal (public) institutional technical education.
This domesday should include:
- Analog circuits
- RF frequency issues
- DSP techniques
- Some statistics tools
- Software craftmanship basics
- Electronics craftmanship basics
- Plenty source of raw data
- Plenty source of annotated technical manuals
- Mad scientific method
There are PCIe to regular PCI 32 bit adapters too
There is some other control signals present in the machine that can be logged along with the linear audio as the servo control track.
These are low frequency, maybe a capturing the servo track in one stereo channel and another time reference signal from the machine such as the head switch signal could help to the dejitter part.
Does anyone knows how the vhs-decode / ld-decode does the sync thing?
In the current gnuradio setup I have, I successfully lock the horizontal frequency, but having some trouble with the vertical.
Also, the luma signal present in the recording doesn't have the same timing as the expected in broadcast / CVBS
It seems to miss half scanline between fields.
A second order bandpass over the luma with a start frequency of Fh and an end frequency of Fh*3 does the magic to provide the horizontal reference.
A trigger / level comparator provides the reconstruction.
The same applies to vertical frequency.
For speed sake, I resample the signal to sample_rate/10 prior to sending it to the filter stage.
Something like audio samplerate of 192kHz will also work
LTO tape seems to have the same width size as in VHS (12.65 mm), I don't know about the formulation, it must have a coercivity between 600 and 800 oersted (for S-VHS).
The thicknes is less in LTO (6.6µm) vs (19 µm) on S-VHS (except E-240).
Fujifilm America still has a section on the web dedicated to the VHS pancake (9 track computer storage medium seems to be tape compatible)
The cassette enclosure and the tape spools could be 3D printed.
The another thing is the heads.
Could hard disk or other type of head be mounted in a custom made device to scan the tape to reconstruct a "magnetic microscopy" image of the recording for further processing?
I made this youtube playlist for this project:
This research is going wild.
The analog format in my country is a variation of PAL, it is called PAL-NC. It not even conforms to the PAL-N specs and it is broadly undocumented.
Last edited by VideoMem; 20th Oct 2020 at 07:33.
Would this be a good choice for LPF? The pass band ends at F1 8MHz, transition band is F2 9.2MHz, and stop band frequency F3 ranges from 12.5MHz-16.5MHz.
Last edited by Titan_91; 29th Oct 2020 at 09:34.