EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: wanton789 on August 21, 2013, 12:12:18 am
-
A board I am currently designing simply involves enabling and disabling of solid state relays. The signals being passed through these relays (i.e. the "load") are each of the 19 HDMI signals (with the exception of the shields and no connect).
On the board:
- The single-ended traces are all routed single-ended..ly, and differential... differentially.
- I have avoided using any vias along the signal paths.
- All traces are very close to ~150mm in length.
- Using this differential microstrip impedance calculator: http://www.mantaro.com/resources/impedance_calculator.htm#differential_microstrip2_impedance (http://www.mantaro.com/resources/impedance_calculator.htm#differential_microstrip2_impedance)
I get a differential impedance of 107.446 ohms.
Currently, the solid state relays (SSRs) I am using are AQY221R6T's from Panasonic. http://www.mouser.com/ProductDetail/Panasonic-Electric-Works/AQY221R6T/?qs=sGAEpiMZZMsUriz2CNI3EyCrq9r2u2G6NbW%252bttKbdKc%3d (http://www.mouser.com/ProductDetail/Panasonic-Electric-Works/AQY221R6T/?qs=sGAEpiMZZMsUriz2CNI3EyCrq9r2u2G6NbW%252bttKbdKc%3d)
Not only is there a limited supply but these are also very expensive.
I chose them because they have a small on state resistance ~0.3 ohms.
I attached a screenshot of the board layout (please excuse the half-completed board, I have removed a bunch of components as our design has changed). As far as the HDMI traces, this was my final design using this relay. Although all of that will have to change if I pick a new relay.
Are these relays necessary? Is my concern for a low on-state resistance to avoid an impedance change even valid?
Off the top of your head, what are some rules of thumb that you follow for signal integrity? (am I even correct in calling this issue signal integrity?)
I would like to pick a much cheaper relay in order to make this board worth the cost.
Please let me know if I can provide you with any more information. All comments, suggestions, criticisms welcome.
-
Relays are not buffers or transeivers I am not sure what do they have to do with signal integrity? Did I miss your question?
Would bus transeiver with tri-state output do better job here? This is complete hdmi transeiver ($37 on DigiKey) http://www.analog.com/static/imported-files/data_sheets/ADV7623.pdf (http://www.analog.com/static/imported-files/data_sheets/ADV7623.pdf).
I just Googled it up actually ;D
-
It's not the resistance of the relay's which would be the problem, it's the unknown impedance of the relays. The data sheet does not show any characterization data for their use as high speed components. Unless they present the same desired 100 ohm impedance they will introduce a discontinuity on the transmission line which will directly translate into reflections.
You could use a network analyzer to extract their 2-port Z parameters and then simulate this in hspice (or similar) or ask the application engineer at that company if they have such models.
Even more problematically the relays are in the middle of the transmission line.... if you need to put something in which has a discontinuity it's generally better to lump those discontinuities together (i.e. get the relays as close as possible to one of those other components)... Are your traces length matched? Not sure what speed of HDMI you are running but it's always good to check that as well.
These don't look like "RF" relays, but rather they are for power applications. They might work, however more engineering calculation and component characterization is needed.
-
You could use a network analyzer to extract their 2-port Z parameters and then simulate this in hspice (or similar) or ask the application engineer at that company if they have such models.
I think usually engineers use an oscilloscope (with TDR function if they are that reach :D) and look at the eye of the signal. HDMI is a digital interface - I am not sure if network analyzer will be a good tool here besides how do you connect it to the tracks?
Some guide from TI: http://e2e.ti.com/cfs-file.ashx/__key/telligent-evolution-components-attachments/00-138-01-00-00-10-65-80/Texas-Instruments-HDMI-Design-Guide.pdf (http://e2e.ti.com/cfs-file.ashx/__key/telligent-evolution-components-attachments/00-138-01-00-00-10-65-80/Texas-Instruments-HDMI-Design-Guide.pdf)
-
with a network analyser one would nornally build a test fixture just for the component parameter extraction. even in tdr using some thing like tthe lecroy sparq would need a fixture.
However thats all academic as the OP has selected a relay which is probably not the right part. He did not describe why he needed a relay... there are probably stronger design choices unless there is a really strange system level problem that needs to be solved.
-
If you really had to use relays, at least use RF relays. TMDS pairs are pretty high bandwidth (1ghz+). Either NEC/Panasonic make one, avialable on digikey for $10.
But what I'd do is probably use a tree topology of solid state HDMI muxers. Depending on how many you can cascade before it gets too lossy, you may need to insert a TMDS redriver. Check out Pericom's selection.
-
It's not the resistance of the relay's which would be the problem, it's the unknown impedance of the relays.
Even more problematically the relays are in the middle of the transmission line.... if you need to put something in which has a discontinuity it's generally better to lump those discontinuities together (i.e. get the relays as close as possible to one of those other components)... Are your traces length matched? Not sure what speed of HDMI you are running but it's always good to check that as well.
Good points, and, yes thank you for the obvious which I should know: we care about impedance, we are talking about AC here (although actually HDMI is digital). Side note: Can digital and analog be thought of as similar in terms of matching impedance? i.e. They are affected similarly by changes in capacitance, etc? Further, they can not be thought of as similar in terms of what other view points?
If you really had to use relays, at least use RF relays. TMDS pairs are pretty high bandwidth (1ghz+). Either NEC/Panasonic make one, avialable on digikey for $10.
But what I'd do is probably use a tree topology of solid state HDMI muxers. Depending on how many you can cascade before it gets too lossy, you may need to insert a TMDS redriver. Check out Pericom's selection.
I believe... you just solved my entire problem.. with one IC. Also, potentially a much more graceful solution to another board.
I3HDMI101 1:1 active ReDriver™. See page 18, 19. http://www.pericom.com/assets/App-Note-Files/AN211.pdf (http://www.pericom.com/assets/App-Note-Files/AN211.pdf)
Game over.
The idea for relays was to switch them on and off repeatedly for some boring automation IOP testing stuff.
Anyways... I am humbled once again by my lack of knowledge... there is so much to learn.
-
I'd definitely go with a solid state buffer IC rather than a relay. You've done everything right in terms of impedance and length matching, but the one factor I've not seen mentioned is capacitance. A solid state relay designed for a very low resistance is probably quite a physically large die, and that means it could easily have enough capacitance to be a problem.
Same goes for trying to probe the signal with a conventional, passive oscilloscope probe - the probe's capacitance will have a fairly dramatic effect on the signal, so whatever you see on the scope screen won't be an accurate image of the true waveform. Active probes and multi-GHz scopes are the order of the day.
-
I'd definitely go with a solid state buffer IC rather than a relay. You've done everything right in terms of impedance and length matching, but the one factor I've not seen mentioned is capacitance. A solid state relay designed for a very low resistance is probably quite a physically large die, and that means it could easily have enough capacitance to be a problem.
Same goes for trying to probe the signal with a conventional, passive oscilloscope probe - the probe's capacitance will have a fairly dramatic effect on the signal, so whatever you see on the scope screen won't be an accurate image of the true waveform. Active probes and multi-GHz scopes are the order of the day.
Active probes! Some day I will have an active probe...
The SSR was pretty tiny... that's why they were $13 a piece. Possible it probably didn't have a lot of cap, but good thing to remember though.
-
The AQY221R6T has up to 100 pF output capacitance (typical 37.5 pF) according to the datasheet. That would have meant a rather low impedance for the HDMI signals no matter whether the relays where on or off :)
The correct solution is, as said before, an active driver with enable or a video switch/MUX IC (or expensive mechanical HF relays if high voltage isolation is necessary).